

AIエンジニアとして採用されるには|Tech With Tim ティム・ルスシカ氏
DataFramed
要約
DataFramedに、YouTubeチャンネル「Tech With Tim」創設者のTim Ruscica氏が登場し、AIエンジニア採用の実情を語った。AIエンジニアの実態は、ソフトウェアエンジニアリングの力にAI特有のスキルを少し足した職種だという。ポートフォリオの作り方、資格の位置づけ、面接の流れ、入社後の評価、コミュニケーションや自己責任のマインドセットまで幅広く議論した。
この回のポイント
- ●AIエンジニアは、数学中心の研究寄りの役割と、既存モデルをアプリに組み込む実務寄りの役割に大別され、後者が大半だとTim氏は述べている。
- ●求められるのは中堅ソフトウェアエンジニア相当のスキル(システム設計、API、Python、Docker、Kubernetes)に、RAGやモデル選定などのAI要素を加えたもの。
- ●AIでコードを書く場合も、小さな単位で生成して自分でレビューできることが重要。コードを読み書きし失敗を重ねることでレビュー力が育つとしている。
- ●ポートフォリオは、デプロイ済みで視覚的に分かりやすく、明確な課題を解くものにする。実行が面倒なGitHubリンクだけでは見られないと指摘した。
- ●資格はあれば有利だが、目的に合う2〜3個で十分。20個集めても意味は薄く、実践的スキルの方が重視される。
- ●面接は、リクルーターとの面談、技術的な質問、実践的な課題の順に進むことが多い。LLM呼び出しをラップするAPIの実装などが出題され、AI利用を認めつつ結果の説明を求める例もある。
章立て
AIエンジニアの種類
数学中心の研究寄りの役割と、既存モデルをアプリに統合する実務寄りの役割という二つの側面を整理する。
必要な技術スキルとコーディング
システム設計、API、Python、コンテナ技術に加え、RAGやMCPなどのAI要素を解説。AI時代のコーディングとレビューの考え方も語る。
ドメイン知識とコミュニケーション
明確で簡潔な文章や会議での伝え方、Gitの重要性など、AI時代に差がつくソフトスキルを議論する。
ポートフォリオとOSS貢献
分かりやすく動かせる成果物の作り方を、AI Plays Flappy Birdの例で説明する。OSS貢献の難しさにも触れる。
資格の価値
資格は補助的な要素で、目的に合う2〜3個に絞るべきだと述べる。キャリア初期ほど有効という見方も示す。
採用面接の流れ
リクルーター面談から技術面接、API実装課題、コードベースのバグ探しまで、受講生の事例を紹介する。
入社後の成功と自己アピール
自律的に意思決定する力や最新動向の把握、期待値の明確化、自分の成果を伝える重要性を語る。
最後の助言と注目の発信者
結果は自己責任で行動し続けることが鍵だと強調。注目するYouTubeクリエイターも紹介する。
解説記事
DataFramedの第387回では、YouTubeチャンネル「Tech With Tim」を運営するTim Ruscica氏が、AIエンジニアとして採用されるための現実的な道筋を語った。Microsoft勤務を経て、開発者の転職支援プログラム「DevLaunch」も運営した同氏が、その経験をもとに述べた内容を整理する。
AIエンジニアは「二層」の職種
Tim氏はAIエンジニアという肩書きが曖昧だと指摘し、大きく二つに分けた。一つはLLMの構築やMLモデルの開発といった、数学に強い博士レベルの仕事。もう一つは、すでに作られたモデルをHugging Faceなどから使い、既存アプリに組み込む実務寄りの仕事で、現在のAIエンジニアの大半はこちらだという。
後者の仕事は、デプロイ、スケーラビリティ、APIといった要素をつなぎ合わせる作業が中心になる。同氏は「ソフトウェアエンジニアがAIを1割ほど扱う仕事」だとも表現している。求人要件の8〜9割はDocker、Kubernetes、システム設計、Pythonなど従来のスキルだったと述べ、博士号や高度な数学がなくても応募できるのに、自分には無理だと諦める人が多いと指摘した。
コードはAIで書いても「読める」ことが条件
コーディングについて、同氏は3年前なら構文まで覚えるべきだと答えただろうが、今は変わりつつあるという。自身もほぼすべてのコードをAIで書くが、小さな単位に分けて生成し、レビューできる状態を保っている。ネット接続なしでコードを読み、何をしているか説明できるかが同氏のテストだ。
レビュー力を養う方法は、多く読み、多く書き、失敗を早く経験することだという。AIは自信満々に誤りを出すことがあり、初心者には見抜けない。経験のある開発者と働く環境に身を置くことも勧めた。
差がつくのはコミュニケーション
AIがコードを書く時代には、人との働き方が差別化要因になるというのが同氏の見方だ。報告はヘッダーや箇条書きで構造化し、日付は曖昧さのない書き方にする。メールの末尾には「自分に何をしてほしいか」を書き、会議ではメモを用意して要点だけ話す。GitやGitHubのブランチ操作、競合解決も、変化が速い今こそ重要だとした。
ポートフォリオ、資格、面接
ポートフォリオの第一原則は「誰でも簡単に理解でき、アクセスできること」である。Docker環境やAPIキーの設定が必要なGitHubリンクは見られないと述べ、可視化やアニメーションなど視覚的な層を付けることを勧めた。課題設定は明快にし、応募先が求める技術(例えばDockerやKubernetes)を使ってデプロイする。自身の例として、NEATアルゴリズムでFlappy Birdを学習させる「AI Plays Flappy Bird」が、数年後の面接でも話題になったと紹介した。
資格については、あれば有利だが決定打ではないという立場だ。AWSなど特定企業を狙うなら関連資格は有効だが、目的に合う2〜3個で十分で、20個集めるのは行き過ぎだとした。特に若手のうちに効果が高いという見方にも同意している。
面接は、20〜30分のリクルーター面談、ポートフォリオを題材にした技術面接、実践的な課題という流れが典型的だったという。大手テックではLeetCode形式が多い一方、一般企業では実践的な課題が目立った。受講生の例では、LLMエンドポイントをラップするAPIの設計(キャッシュ、レート制限、ポーリング応答)が課され、続いて既存コードベースのバグ探しが行われた。RAGとファインチューニングの使い分けのような質問もあり、AIの利用を認めつつ出力の理解を問う形式も見られるという。司会のRichie氏も、DataCampの採用で同様のやり方をしていると述べた。
入社後と最後のメッセージ
AIエンジニアは中堅として採用されることが多く、1〜2か月の立ち上がり後は、自律的に判断し、設計し、必要なら若手に仕事を振る力が期待されるという。最新のモデル動向を追い、選択を説明できることも評価に関わる。また、成果を適切に伝え、最初の数か月の期待値を上司と確認しておくことで、後の昇給交渉などの根拠になると助言した。
結びでは、結果は自分の行動次第だという厳しいメッセージを送った。大量に応募する、ユーザーのいるアプリを作る、LinkedInで発信し続けるといった行動を挙げ、1〜2年続けた人は仕事を得ているという。
まとめ
この回の核心は、AIエンジニアが特別な職種ではなく、堅実なソフトウェア工学にAI統合の知識を足したものだという点にある。日本の読者にとっては、既存のバックエンド経験を活かしてRAGやLLM APIの扱いを学び、動くデモを公開する、という現実的な戦略が見えてくる。AI生成コードをレビューし説明できる力や、簡潔に伝える力が評価軸になるという指摘も、転職準備の参考になるだろう。
文字起こし(英語・自動生成)
Generative AI is transforming industries at an unprecedented pace. But as AI changes how you or your team work, one thing is clear. Your skills also need to evolve. At Datacamp, we offer everything you or your team need to adapt and thrive with AI. Whether it's business users looking to get the most out of ChatGPT and Copilot, or developers and data scientists looking to fine-tune models, you can learn the entire AI skills spectrum on Datacamp. How are your AI transformation today? Start learning at datacamp.com I'm telling you, if you apply to 20,000 companies, someone's going to call you back. I'm telling you, if you build a really cool app that gets 100 users, somebody's going to notice. That's the biggest point I can give, because most portfolios never even get viewed because they don't have that, right? Because it's really just software engineer who happens to do like 10% of work with AI. That's what I see. AI engineer is the hottest job title right now. So today we're going to find out how to get hired in that role. I want to know what skills and qualifications you need and what makes an AI engineer successful.
So a lot of people actually thought that they were excluded from applying to AI engineering roles because they didn't have a PhD or they didn't know all the advanced math. When realistically, all these companies really wanted was a few AI specific skills. All the people who just said, oh, it's too hard. Where are they right now? They're back where they were at the beginning. Our guest is Tim Rusica, more commonly known as Tech with Tim. He's a software engineer turned educator and entrepreneur. Tim runs a technology education YouTube channel with more than 2 million followers. He also runs kernel management, helping other content creators turn their content into businesses. Let's find out how to become AI engineers. Hi Tim, welcome to the show. Thanks very much for having me. Excited to chat. Yeah, great to have you here. so AI engineering is maybe the hottest role around at the moment but it feels like it's lots of different roles kind of smushed into one job title so do you want to talk with you like what are the different facets of being an AI engineer yeah it's funny like AI engineer can mean so many different things depending on where you're working or who you ask so I would say that
you know okay maybe we can break it into like two or three different categories I would say one of them is going to be more your math heavy, kind of more PhD level, doing like actual, you know, building an LLM, right, or writing ML models or doing like all this kind of fancy technical stuff that, to be honest, probably neither of us really know how to do. And most people are not doing commonly unless they went to school for this. And then you have the much more practical side where you're taking all of that work that's been done by kind of the geniuses of the world and integrating it into applications. And I would say that's where most AI engineers are lying nowadays, where you're using an LLM, you're using a model, you're using a hugging face transformer, whatever, and you're building it into an app that already exists. And actually what that means, I'm sure we're going to dive into that, is that a lot of the work you're doing isn't really like raw AI ML related stuff. It's more kind of gluing the pieces together and knowing some of the concepts and kind of how they play, which gets into deployment, scalability, APIs.
I'm sure we'll dive more into it, but that's kind of how I view the field right now, like super technical, lower level math stuff, and then more of kind of integrating and that layer on top. Absolutely. And these feel like very different roles then. So, I mean, one of them is like, I guess, more closer to the traditional machine learning background, but maybe at a larger scale. I think the one's closer to software engineering. So on that note, since AI engineer is quite a new role, people are generally transitioning into AI engineering from somewhere else. Do you want to talk with me about what background you need to get a job as an AI engineer? For sure. So I've done a lot of videos on this. Again, I think the term is so vague that obviously do your research and see what role you're targeting, what companies you're looking at, because they're all going to have their own requirements. But generally speaking, it's going to be the exact same list of stuff you would normally need for like a mid-level software engineering rule plus some of those AI concepts. So we'll go through some of them. For example, system design, right? Okay, how do you design an API? How do you have maybe some basic load balance
or some caching? How do you think about kind of scaling a system up that gets a little bit larger, especially when we're talking about inference of things like LLMs, right? And then we have kind of the AI layer where we're talking about, okay, how do we pick which model to use? How do we verify that we're getting the correct result? How do we rate limit something to make sure we're not going through too much? How do we manage the context or set things like RAG, right? Retrieval augmented generation or connecting to external tools, adding MCP servers. Again, it really depends what it is that you're building, but it's kind of take all that traditional software engineering stuff, add that practical AI layer in terms of how do I call these APIs? How do I work with LLMs? How do I get the response back in a correct format and combine those two skills together? And then, of course, build it, demonstrate to an employer that you actually know what you're talking about. And then I would say, to get a little bit more specific, a lot of that now has to do with using Python, for example. I mean, most of your LLM wrappers, SDKs and stuff are going to be in that language, so being really comfortable there. And then being familiar with maybe some of the frameworks,
depending on what you're using. And even Docker, Kubernetes, all those things are still equally as important when you get into the era where you're calling LLMs thousands of times per second, depending on the app you're building. It is basically, it's all your standard software development skills and then a few extra ones on top of that. It's kind of interesting that Python got mentioned towards the end. There's a lot of open-ended questions about, like, is coding still worth it? Like, do you need to have coding skills? Do you want to talk me through, like, what your position is on, like, what do you need to know about programming? Yeah, it's funny because if you were to ask this to me, you know, three years ago, I would have said, yeah, you need to know every piece of syntax. You need to be, like, experts still in your domain, know the language. And while I still mostly believe that, it's definitely shifting pretty quickly, especially as I see even people that I know, they're in this industry that don't actually have any real practical coding skills that are building like incredible applications. So I would say if you want to just if you want to be a software engineer, you want to work for a large company, absolutely, you should still know the fundamentals of programming. Does that mean you have to memorize every single piece of syntax?
Does it mean you have to write all of this code by hand? Absolutely not. And I think you can now skip maybe a lot of things that I spent a bunch of time learning back, you know, 10 years ago and really accelerate the journey. But ultimately, my kind of test is like if I gave you a piece of code and you weren't connected to the Internet, could you read through it and explain to me what the code is doing? So there's a lot of things that you do need to know in order to do that. And there's some things that now you can kind of skip. You don't need to know, you know, the exact method name and all the parameters and things that maybe before would have actually helped you speed up and write code a lot faster as a developer. So that's kind of where I'm standing. I would say now, and even myself, pretty much all the code I write is with AI. I try to do it in kind of smaller pieces so I can actually review it and kind of have, you know, still be in the driver's seat in terms of what's going on in the high level decisions. And it's tricky because if I were to just to recommend, hey, do everything with AI, people get this massive mess that's not maintainable in the future. So it's a really weird place that we're at right now. Yeah, and it's really tricky to navigate. And I don't, it's kind of always difficult for me to suggest to beginners, what should
you do nowadays? I know that's not the question, but something I'm always raveling with in my mind, like what is software development going to look like even six months from now? And maybe this answer is already going to be outdated. So who knows? Yeah, it's certainly changing very quickly. And like even a year ago, it was very unreliable in terms of like generating code. But it's got to the point where it's like, oh, yeah, you can actually build something that's pretty good with minimal intervention. So I guess related to this is do you need to be able to review the code that has been generated by the AI? Like what sort of skills you need to be able to do that effectively? Yeah, I mean, to answer that, yes, absolutely. You should be reviewing it. I think, first of all, there's a skill in terms of how you get the AI to generate the code and also how much of it you generate at once, because it's just impossible even for experienced software engineers to review 20,000 line pull requests, right? Your mind just can't grasp that much information and review it. So I would say, first of all, the first skill is like, how do we get AI to make small, targeted changes that do save us time, but actually keep saving us time because it doesn't
require us 20 hours to review. So that's kind of the first side of it, how you do that correctly. And the second side of it, honestly, it's kind of a, I don't know, maybe a bad answer, but the way you get experience reviewing code is reviewing code, right? And how do you do that? You need to just read a lot of code, write a lot of code, and also see a lot of mistakes that you've made yourself or other developers have made so that you don't make those again in the future. The only reason I know that this function is poorly written is because I've written a function like that before, right? And I've seen what that actually does when I deploy it out or I've had my mentor at Microsoft when I work there rip it apart and tell me, you know, never do that again. So there are a lot of learnings that you really only can get by making those mistakes. So how do you do that? You just got to make them as fast as possible, right? Push out a lot of code, build a lot of things, see where stuff breaks and review all of these AI decisions and kind of realize, hey, this isn't just God. It doesn't know every single answer. And sometimes you're actually correct and you've got to kind of guide the AI. And I've had AI make some crazy things that it's 100% confident. And I'm thinking like, okay, if I was a complete beginner, there's no way I would know this
mistake. But I know from my experience, like we can't do this. Let's fix that and let's go back. So I would say just kind of racing to get that experience as quickly as possible and put yourself in an environment where you're working with other developers who already have the experience so they can kind of rip apart all the stuff that you're doing that you might not already know. Right. yeah i mean i love that idea of just trying lots of things just building lots of things and then having that feedback loop of like okay this broke and how can i do things better i just raised it i mean so we talked about some technical skills and one thing that's commonly suggested is you need some domain skills as well like what what does that mean and what might need to learn yeah so i mean when i think of domain skills i think of the maybe auxiliary skills that aren't specifically software engineering so that could be even how do you communicate effectively like as an engineer with other people? How do you send a Slack message that's actually going to get your point across, right? Or if we talking about like a specific industry let say you building you an AI engineer for like a robotics company you probably need to know a little bit about robotics right Or what the hardware stack they using Or what the goal you actually trying to achieve
So I think it's super vague. And we'd have to kind of narrow down on like, what role are we talking about? But it's everything outside of that. And I would say now, especially those social skills, the way you work with people, the way you communicate, even how you formulate an email, there's these small little things that really do make you stand out as a developer, especially as the coding stuff is becoming kind of more redundant and done a lot by AI. And just the way you work with people, I mean, that's the thing I noticed right away when I work with anyone. How do they communicate? How clear are they? What do they ask for? What kind of questions do they ask me? Do they have a deep understanding of the problem? So that's what I would focus on. Outside of, again, there's small little details you'll need to know about any industry, right, depending on what you're writing the code for. Absolutely. I feel like you've just terrified a lot of people. like oh social skills i'm just i want to talk to the ai not to other humans but it is incredibly important like you're always me working as part of a team so yeah those sort of uh team working skills do seem uh incredibly uh essentially in in most roles okay uh so yeah i like the idea of having uh like some uh robotic skills be working robotics or whatever you need to understand what
what the business is you're in are there any more soft skills that you think are useful to have Yeah, I think, again, I always just go back to communication because ultimately what sets you apart from an AI model is what you say, right? The way you communicate, like the fact that your little box pops up in the Slack message and says something non-AI model and people trust you more because of who you are, right? The fact that you're actually a human being. So I would say really just focusing on being really clear in the way that you communicate. I've been in a lot of like engineering teams where it's very messy. There's people aren't very clear. They're very vague where we spend 10 minutes just trying to understand what questions someone's asking or what the bug was or what's actually going on. And just having kind of a logical structure, like thinking process when it comes to even giving an update. Right. Or asking a question or, you know, describing a blocker or how you communicate on like a stand up. Right. Someone comes to you. Hey, Tim, can you give me an update on, you know, what you worked on last week? Having that kind of prepped, ready to go and just giving the important information that someone needs to know to proceed.
It's hard for me to go into many other skills. Again, depends like what it is you're working on. I guess a small one I could say is maybe not soft skill, but how you work with like Git and GitHub. I think especially now is like super, super important. We have so many rapid changes happening all the time. So especially now understanding branches, being able to revert back to something, merge conflicts, that's one that I've seen become more and more important in that while AI can do some of that, you've still got to know a lot of those commands and how you don't completely mess up the work tree that you're on. So I guess I'd throw that out as kind of a bonus there. Absolutely, yeah. Git and GitHub or the sort of version control tools is incredibly important for anything if you're writing code. Also, you mentioned just how to communicate within a standard meeting. like meeting skills in general i think this is broadly applicable even beyond like whether or not you want to an air want to be an ai engineer so yeah do you want to talk me through like what are your top tips for doing well in meetings yeah i mean again it's so like specific as it depends where you're working i would just say being clear and unambiguous and it's funny because i remember
when i was in university i actually took a class like technical writing for engineers and i used to think and most people still do right universities kind of use this you're learning a bunch of stuff you're never going to use. But the one takeaway I had from that class that I still try to apply today is writing things and speaking in such a way that there's no way that someone could misinterpret or misunderstand you, right? And you just say something so clearly where you've actually really thought through the way that you're going to describe it and the way you're going to discuss it so you can just get the point across quickly and you don't have any potential mistakes or mix-ups based on kind of the way you've communicated something. Kind of blanking on a great example for you right here on this call. But even the way that I would write an update, for example, in a Slack channel, which now is happening all the time, I would start with like a quick header. Hey, Tim's update for this week, right? And put like 0731. But even rather than that, I might do something like, oh, July 31st, just so it's clear and we don't have any mix up on day, month, year, whatever. Then I'd have a quick bullet point like, hey, this is a quick
summary of whatever I'm talking about. Some action items, questions, blockers, and put like subheadings, formatting, just so it's really easy for someone to parse through it and read it and just try to be consistent with that format, even if it's just the way that I speak so that they know, hey, this is what Tim's update is going to look like. This is how I can parse it. This is what I can expect in the future. Even if you're writing an email at the end of the email, having like what I need from you, three quick things just so they can quickly go to that and give you the answer and making things like as easy for people as possible. In a meeting, I would say have your own little notes on the side, notepad file and a cloud code session, whatever. Hey, this is my update. You can literally just read off of that or at least have some bullet points that you want to hit. So you're not rambling kind of like I am on this answer. And you give, you know, the direct point because in meetings, everyone just sits there and want to talk forever. Let's just get through the important stuff. In these 30 minute meetings, we can make five minutes if everyone comes prepped with what it is they want to share. Right? No, I love that. Everyone's just deluged by information. So being concise and being kind to the people who are receiving your information so you keep it short and clear is uh
yeah that's gonna take you a long way i think as a skill i think it will just make you stand out a lot of people just send these huge big paragraphs and it's kind of like what's the point give me the tldr like what do you need from me right so i like to just lead with that and if there's other details someone else wants to read fine but ultimately i just want to move things as quickly as possible so that's typically the way i communicate that i find most people appreciate Yeah, I definitely have colleagues where sometimes I'll copy and paste that Slack message into AI and it's like, can you summarize this message for me, please? Yeah, that's one of the great uses. Yeah, less rambling from your colleagues. All right. So before you mentioned that just building a lot of stuff is incredibly useful for getting hired. I suppose related to this is the idea that you can have a portfolio to show off your work. Do you have a sense of what goes into a good AI engineer portfolio? Sure. So portfolio is a difficult thing because there's, especially in AI engineering, there's a lot of really interesting stuff you can build. It doesn't really present that way, right, if someone's going to look at it.
So the first rule I typically have with a portfolio is it needs to be like dead simple to understand and to access. I was hiring software engineers before and even like AI adjacent guys, and they would send me like just a GitHub link. And I'd have no idea how do I run it? How do I view it? Is the project up here? what's actually going on. It's like they expected me to like spin up a Docker container on my own computer and get a deployment of a VPS and have a MongoDB database and put in some keys and an ENV file. Like if you have anything like that, no one's ever going to view it or ever going to run it. And the most it's going to do for you is just be a talking point if you actually get into an interview, right? So the first thing with portfolio is just make sure whatever it is that you're building, build some kind of visual layer and interface on it. Even if it's just like an animation showing what's going on in the backend so that someone can really quickly understand what it is that you've actually done. That's the biggest point I can give because most portfolios never even get viewed because they don't have that, right? The second side of things is I would just make sure that whatever you do solves a clear problem.
And again, just makes sense to the interviewer. If you do some really random, obscure, crazy backend thing and it takes you five minutes to explain it, it doesn't matter how complicated it is, they're just not gonna get the point. So for example, one of my best projects I had was actually kind of an AI project was AI Plays Flappy Bird. That's literally what I called the project. It's one of my most viewed videos on YouTube. And if you read that on my resume, you're like, oh, I get it. It's an AI that learns how to play Flappy Bird. And then I just added a few details. Oh, I used the neat algorithm in Python to build a self-evolution loop of AI models that evolved it and learned how to play the game. And you get it right away. And you can view it and you can see the cool animation and it's a graphic. and that got like a lot of looks and a lot of questions when I was interviewing, even though that was like five or six years ago. So from there, I would then say just make sure you have some of those kind of big keywords right in the description so they get it. Hey, integrated with OpenAI API, connected with LandGraph to build a simple AI agent that handles emails for 1,000 people, you know, whatever, right?
Like something like that. And then the rest of the portfolio project, you can kind of explain and will be a talking point that you can go into in the interviews because generally speaking, the portfolio might kind of get you in the door. It might like get someone interested. They're not going to dig through and read all the GitHub code. It's going to then be up to you to kind of explain it, talk about the challenges, the mistakes, you get the idea. So make sure it's deployed. Make sure it's simple to understand. Make sure it's visual. And then, of course, try to build something that's unique that not a million other people have built with some of the keywords, languages, and technologies that that company is going to be hiring for. If they're asking you need Docker and Kubernetes, deploy it with Docker, right? Have a Kubernetes cluster set up, and then you can talk about that and prove that you have those competencies for the company. I love that idea of just making it really simple for people to understand. I think a lot of hiring managers, like the whole point of looking at a portfolio is just to check that you're not a moron, but you do have the skills you claim to have. So they just want to do that as quickly as possible. So beyond creating your own projects, do you think there's a role for contributing to an open source project or contributing to other pieces of software that are maybe not owned by you?
Absolutely. I mean, that's obviously a great thing to do and is going to make you stand out from like 99% of people because most people just won't do that. That said, it's not easy to do that, right? And most of your contributions are going to be super small. You're probably not going to be like a lead contributor on some, you know, random open source project that's out there. So while if you can do that, you know, I would recommend it. To be honest, I don't have that much experience working on that. I don't even really know how I would guide you on which project to work on or how you could get involved with that. if you can figure it out great but I do acknowledge that that's like fairly challenging to kind of just oh yeah go contribute to an open source project and probably there's some other steps you could do that might bring you a little bit further like if you had three kick-ass portfolio projects that were super clear deployed had some real users for example and that was shown on your what is it on your resume or whatever I would imagine that would stand out a lot more than like made a 40 pull request for VS Code or something right as an open source project. Absolutely, yeah. So, generally, you've got to be, like, using the open source project, like
the software for a long time before you're like, oh, yeah, maybe I can contribute to this. Okay, so, related to the idea of a portfolio is qualifications. So, what sort of qualifications do you need to become an AI engineer? For sure. So when it comes to qualifications, I think that these can be useful and they can show, first of all, you put some effort into something that you've gone through a program that you're like kind of more committed because you've actually gone through with the process. OK, I'm going to probably pay for this, go spend the time doing it. But generally speaking, companies will probably still value your ability to demonstrate those skills and prove them over the qualification. That said, for certain roles, there's like very specific qualifications that companies will look for. Right. If you want to go work at AWS, probably it's a good idea that you've completed some of the AWS certifications, right, that are kind of industry known. You can go through, you get the idea. If you want to be an AI engineer, of course, it should be related to AI engineering. You should probably have maybe a Python certificate. You should have something like LLM Ops. You might have calling LLM APIs. Obviously, I know we're with Datacamp here.
They've got a bunch of great certifications with their courses that I recommend all the time. You got to look at what role you specifically want to see which ones are going to make the most amount of sense. And then for kind of some very specific AI engineering roles and other dev roles, there's a whole kind of heap of them that apply for like, hey, C++ with Microsoft, right? And they've got their .NET, whatever line. You've got Google Cloud certifications. You've got Azure certifications. But I wouldn't go collecting a ton of them. If you get just the ones you think are genuinely useful and that you need, stop there and then just try to build some real practical skills because that's going to bring you further. So if you tell me you've got 20 certifications, okay, probably we've gone down a little bit too far. two or three that makes a bit more sense to me right yeah so it seems like all the big tech companies um you mentioned to have their own like extensive suite of certification then of course 30 camp has some as as well so thank you for the mention but yeah i can certainly say i have 20 certifications that's not going to help you over having like two or three that are targeted to the job that you want to do exactly and it's a nice to have but it's not if you don't have one don't feel bad it's not going to like rule you out it's kind of like oh okay i see this guy's
put some effort in oh specifically aws okay at least that proves this point okay he's got some experience with llans okay that's nice to have but it's maybe like two or three percent of like you know that hundred percent picture that's going to get you in get you hired get you an interview right absolutely and i get the sense that the certifications generally work a lot better like when you're earlier in your career rather than later on yeah a hundred percent okay uh all right so uh tell me what does the hiring process look like like uh what once you get uh you put in your application they say okay uh come for an interview like what what is involved there yeah and this is AI engineering specifically you're talking about, right? Sure. Cool. So I actually ran a program. I showed it down in January, but it's called Dev Launch, where we help developers land developer positions, whether it's their first position or kind of like a better position upgrading from where they're at. And we had a handful of people that actually did get hired as AI engineers or that were working towards that. So I'll just walk you through like practically what I saw and keep in mind things change constantly and I'm not actively interviewing. So I probably don't have all the most up-to-date info, but here are some changes that I noticed
compared to what I would see in other fields. So firstly, a lot of these companies were obviously looking for people that had strong software engineering skills with that AI layer, like I talked about. So a lot of people actually thought that they were excluded from applying to AI engineering roles because they didn't have a PhD or they didn't know all the advanced math, when realistically all these companies really wanted was a few AI-specific skills that really just built on top of strong software engineering skills. And 80 or 90% of the skills that I would see in the requirements list were like Docker, Kubernetes, design systems, Python. You get the idea, big old notation, data structures. So most people already have a lot of the skills. They just need to add kind of that small layer that we touched on before. So assuming that you created your resume in a way where it spoke to you being an engineer, it's specifically tailored to that type of job posting. You had all the keywords and you got called in like some of our students did. What I would typically see is the first round would just be a quick recruiter check, kind of a fit check to make sure they're not going to waste, you know, like an engineer's time interviewing you.
20, 30 minutes, walk me through your resume, explain this gap. OK, why are you interested in this position? And, you know, I'd say 80 percent of people will pass those pretty easily as long as you have some decent social skills and you can kind of explain everything you have and you're not bullshitting. Right. Like what's on your resume from there? You typically get called in to like kind of a first round interview. First round interview, same thing, a little bit of a lighter round where typically they're just going to ask you maybe some technical questions. Hey, tell me about your experience here. Walk me through a portfolio project. Why do you want this role? Are you comfortable with this technology where it's a little bit more challenging? You're going to have to know your technical stuff. But I didn't see that many people getting asked like real technical challenges or like coding problems in kind of that first, let's call it real round. Then after they passed the first round, the second round is where I saw a lot more coding challenges. Now, if we're talking Microsoft, Apple, Amazon, big tech companies, they're always giving you these like leak code style rounds and you're going to get multiple of those. But for more of the practical companies, I would say large enterprises, ones that aren't
just like big tech, they actually had like very practical interviews. So I'll give you an example of what I saw for a second round for one of my students. He was actually asked to implement a call to an LLM and wrap it into like an existing API. So effectively, they had their own, I don't know, some endpoint for some LLM that they wanted to call her some internal model. And they said, hey, here's the endpoints. Here's how it works. I want you to build an API around this that will cache different requests. We'll do some rate limiting. We'll make sure that we have like a polling response available for getting the API callbacks. It can take like one or two minutes for the response. And he was responsible for like designing that API. Right. And fortunately, he did well on that. Made us the next round. And then they asked him, I think, to actually sit with another engineer and like break down an existing code base and try to find a bug in like some AI, again, really just like a system around some AI model that existed. So that's typically what I was seeing. I saw some other people get like, again, those logic puzzle leak coding type stuff, but usually it was actually pretty practical.
And it really was just like core software engineering skills with some endpoint related to like some ML thing. And maybe they asked you like, oh, when would you use RAG versus using fine-tuning or some some quick questions like that so yeah hopefully that gives you a good sense but that's typically what i'm seeing for ai engineering roles i'm sure it changes all the time now answer some theory questions write some code you actually in this case they let you use ai to write the code but they wanted to see how you used it and like if you were reviewing the code and look at what editor and what tooling you were in and yeah that's what i noticed i know that's very cool and actually it just sounded like quite fun problem is like oh yeah you're building this API and figuring out all the system architecture and things like that. Nice. So yeah, that just seems like you've got a mix of you need to understand the basic theory around AI also you need to be able to build stuff as well maybe with AI assistance. Actually, that's quite closely aligned with how a lot of data camp roles are hired for. It's like you can use AI during the takeaway but then you get grilled on whatever the result is in an interview
afterwards so you've got to be sure you understand what was generated okay uh so uh suppose you get the job how is success measured like what's going to happen in the first few months and how do you uh prove that you are worthwhile as an ai engineer yeah i mean i think ai engineers typically are also coming in at like more of a mid-level as opposed to like a junior or an intern i'm sure those roles exist and there's some of them junior engineer whatever but typically it's like hey you should already have three four five years experience you've been a software engineer before and i see a lot of people kind of transitioning to that role. So let's just keep that framing when I explain the, you know, like what's expected. Generally, what I've seen is a lot of companies hiring these are also like fast growing startups. They're maybe not as big, right? They're not building their own AI from scratch. Like the reason they're hiring this role is they want someone to be able to come in, especially if it's a smaller company, and integrate AI into their product, right? So how is success measured? How well are you able to do that, right? Can you provide the impact? can you come in and can you kind of take charge and do you have your own autonomy and can you build
some things without being handheld right are you always constantly going up to uh whatever your mentor senior manager and asking for help or saying you're blocked or xyz that's going to be a pretty quick way i think for them to be looking for someone else so generally what i've seen and with one of my students who landed the role there is it's very it was a lot very hands-off like hey this is the problem we have this is what we need you to take care of sure you know work with other teams ask us questions, but after the first month or two ramp up understanding the code base, it's kind of like, hey, we're looking to you to make a lot of these decisions. And of course, at a bigger company, you're going to have those different layers of bureaucracy, your senior engineers, your managers, where they're probably going to do a lot of the architecture and kind of hand off to you the steps to complete. But because I know that this is a role that is popping up a lot and at smaller orgs where they might only have one, two, three, four AI engineers, you've got to be comfortable kind of deciding what it is to actually build, making decisions, breaking down the plan, maybe handing it off to an intern or someone junior. I think those are skills that a lot of engineers don't have because they're used to working
underneath someone or being told exactly what to do. And in AI, when things are changing so quickly as well, you need to kind of be up to date with the current trend because someone might come to you and say, hey, why are we using GPT-4 Nano? I thought we're at like 5.5 now. If you weren't the one to kind of figure that out and change that over or update the code, I don't know, I don't think that looks that good. So I would say staying up to date in the current industry, right, being able to kind of make decisions on your own and fight for yourself and justify why you decided to do what you did, and just overall provide impact in whatever way that can be. And again, most of the time, that just ends up being like writing a normal software system that happens to have some AI integration that you kind of just knew to use this integration versus another one because of whatever experience you have with AI engineering right Absolutely I mean for all the sort of drama around oh AI is ticking over everything it like yeah it actually a software engineering role with a few little tweaks to it So a lot of the same sort of metrics for success apply there. But also interesting how you talked about the differences in like how bosses portray like the problems to you.
Like I guess, yeah, maybe at a junior level, they tell you how to solve the problem, but at a mid-senior level, it's like, well, here's the problem, go figure it out. and yeah that's the skill you need to learn. A hundred percent and like just having also some of that like system architecture some of that forethought for the future as well like not just these kind of one file solutions like the whole system as a whole and that's what I've noticed with a lot of AI engineers is like really it's just doing a lot of system design that happens to revolve around some kind of LLM or ML model or whatever that's happening in the background so it's almost like just a senior role condensed into this AI one that's kind of around this new topic So it's so funny to me that we call it that because it's really just software engineer who happens to do like 10% of work with AI. That's what I see. Absolutely. So while we're on the subject, have you got any more tips for success at work in general? I would say, you know, sometimes you kind of have to play the game to some extent, right? You have to advocate for yourself. You have to demonstrate the work that you're doing. You have to kind of talk to the right people, know what's going on.
You can't just be in your own little bubble, kind of sitting in the corner, doing some work, whatever, pushing it out. Like at least when I've been successful, work and when I've gotten praise or whatever, it's like, hey, Tim takes initiative. Hey, he asked good questions. Hey, he's organized, structured, you know, speaks clearly, whatever, right? Like kind of a lot of those communication skills come up. But it's also because I'm actively trying to make myself look as good as possible and not just kind of complaining that I'm not appreciated or saying, oh, he doesn't know like all the stuff that I'm doing. I think a lot of people, they don't really like stick up for themselves as much as they can in a workplace environment, not in like a combative way, but they just don't give themselves all the advantages that they should have in order to maybe get a promotion or get praise or get a return offer or whatever the situation may be. Ultimately, what I say is like, no one's going to look out for you as much as you need to look out for yourself, right? No one cares about you as much as you care about yourself. So you've got to do everything in your power to not in a bad way or boastful way, but just make sure people understand what value you bring, the contributions you're making. And also, if you are making decisions or you feel like you're worth more, you have a problem,
like you need to bring that up proactively and kind of set targets. Like even now when I do kind of freelance work, other work, I'll chat with whoever's in charge. Hey, what do you need from me for this to be successful, right? What does it look like in the first three months of like, where would you want me to be at? How quickly are we like expecting this to get done? Like setting the expectations really clearly. So then you have something objective that you can always compare back to. If you do ever make an ask in the future to say, hey, I want a raise or hey, I want a few extra days off or hey, I want this person under me or you know what I mean? like you need to have something to go back to which most people never have and then they make the ask without any metrics and they're kind of starting from square one you know absolutely i mean i think this is a very common problem uh with technical roles i mean certainly uh that resonated a lot with me earlier in my career where it's like oh yeah i'm doing the work but somehow don't feel like people recognize it and it's like oh well i didn't you know i didn't tell people about all the cool stuff i was doing and yeah uh that does seem very important to play the game a little bit it's a balance obviously it's tricky but just advocate for yourself you know what i mean
and and just show people what it is you're actually doing and don't be kind of shy or nervous i know it's easier said than done in the tech space but the more that you can be i don't kind of understand those social dynamics in the workplace because ultimately the better off you're going to be in pretty much any role but especially tech when most people suck at the social side of things if you're even like a five out of ten you already had a most engineers and um you know use that to advantage right oh yeah i mean there's the old joke about like how do you tell uh if a software engineer is extroverted it's like they look at your shoes when they're talking to you instead of their own actually you know one of my one of my tips for like playing the game sometimes i will send a slack message in like a publicly visible channel with like lots of people in it at like 9 p.m or something because it looks like i'm hard at work actually i'm on the sofa like just watching tv and i've just sent it from my mobile phone but it looks like i'm you know concentrating on work things. That's actually a really good point. Sorry to cut you off. But like, if you can also just be quick to reply, like so many people take so long to reply to everything. I mean, for me, I don't I don't have like that much of work life balance, but I'm always on my
phone like constantly. So if you ask me something, even if it's 11, 11pm at night, if I see it, and I can answer it really quickly, I will, I don't set the hard rule of like, Oh, I'm not doing any work after 5pm. Some people might not like that, they might disagree with me. But that's a really easy way to kind of push yourself ahead when people know like, hey, you can get a quick answer from this guy we can move things forward really fast and that's helped me like drastically and pretty much any rule i've had absolutely i mean certainly uh if you're in an organization where it's spread across multiple time zones you're having a lot of companies now it's like yeah just uh being available at a broader set of hours works pretty well all right uh do you have any final tips for how to get a job as an ai engineer i mean it's gonna be it's gonna sound like maybe uh i don't know i mean tip a harsh tip but generally speaking like everything's in your control. You know what I mean? And you kind of just have to have that mindset if you want to be successful at pretty much anything in life. And that is the exact same for getting a job. Like we could sit here on this podcast and we could complain that AI is taking over and it's harder for juniors and college sucks and the industry is bad and people are getting paid less and
inflation's here. Like there's a million things that we could go through, right? And that would be a pretty shitty podcast to listen to. And also just like, it doesn't help anyone in any way. We're just sitting here kind of observing what's going on and complaining. But a lot of people like to put the blame on anyone other than themselves. Oh, well, these companies don't appreciate me or, oh, they don't like my role or, oh, things are changing too quickly. Ultimately, it's all your fault. You're the only one who can help yourself. If you do want to get a job, you need to change something in your approach. You need to take accountability for it, right? I'm telling you, if you apply to 20,000 companies, someone's going to call you back. I'm telling you, if you build a really cool app that gets 100 users, somebody's going to notice. If you post on LinkedIn for six months and share stuff that you're doing, that's going to help you. You know what I mean? There's lots of things you can do to improve your situation. And so many people that are probably a lot dumber than you have gotten tech jobs simply because they put in more work and they didn't complain about the environment. They just put their head down and did what needed to be done. I get it's frustrating. I get it's annoying. It's not easy. It's probably going to take a long amount of time. But anyone who's done this for one or two years, they do get a job. They
do find it. And the reason is because they're actually taking accountability and trying to better themselves and change things in their life. And the worst, worst, worst case scenario is you don't get a tech job, but you find some other job or you find something else and you've proven to yourself that you can do all of this work. It's only going to better you. So that's just generally the mindset you have to have. You can't be a victim. You can't just complain. Things are going to change. Things are going to get tough. And the only people who succeed are the ones that have that kind of self-accountability mindset. And that's just how you have to do it. That's the answer. And that's, I would say, 90% of the battle is purely mental and that it's just doing the work and doing the things that need to be done right oh man that was tough love but a lot of people need to hear that uh and oh so your point about like maybe don't get a tech job actually i mean one of the great things about ai being everywhere is like every single industry needs people have some ai skills so yeah there is more than just like a few tech companies to apply to there's lots of stuff you can do you know i'd be like if you told me tim you need to get a new tech job in the next 30 days and you can't use tech with tim or any of your
branding, like, I would figure it out, you know what I mean? Like, I'm going to do what needs to be done and just go through the work and like not complain about it. And all the people I know, and all the people that I've helped that I told them, hey, do this, this, this, it's a lot of work, it's a pain, it's going to take you a while. Every single one of them that actually did that and didn't give up got a job. Now it took a different amount of time for each of them. And some of them got unlucky and some of them failed interviews and some of them had a weird situation, whatever, but they got through it. But all the people who just said, oh, it's too hard, where they were right Now they're back where they were at the beginning. So that's why I just frame it like it is so mental and I get how frustrating it is, but that's just the situation we're in. And if you want to be in this field, which is still one of the best jobs to be in, some of the best benefits, some of the best work-life balance, some of the coolest stuff you can possibly work on, at least in my opinion, like it's going to be hard and people are going to be competing for it. And it shouldn't be easy. If it was easy, you wouldn't even be fulfilled if you got the job and you probably wouldn't be going for the job because you wouldn't have all the great benefits that you get. Right. I mean, that's that's the way I look at it. Absolutely. Yeah. Yeah, there's a worthwhile prize you've got to put in the effort. Exactly. Nice. All right, to finish up, I always need more people to learn from.
So whose work are you most excited about at the moment? Hmm, that's interesting. I mean, honestly, a lot of what I'm doing right now is just AI content, AI tools. I think there's a lot of interesting people on YouTube, like, that are purely in the AI space because I've kind of, at least for me, gone through a lot of my dev stuff already, and I'm no longer interested in, like, Python tutorials online. so I like guys like Nate Herc David Andres Jack Roberts I believe these are like some of the up and coming AI well I mean they're pretty established like content creators I would say that do some pretty interesting stuff like AI workflows AI coding all kinds of integrations and they're less like pure dev focused and more kind of new era like vibe coding stuff if you know what I mean and then in terms of just the classics one channel I always like to keep up with is called Sendex not sure if you've heard of him but he's a big a python developer he's actually was my inspiration for starting my youtube channel like 10 plus years ago now and he does like really low level like robotics and hardware and building an lm from scratch and building neural networks from scratch and some more like hardcore coding stuff so if you guys like that then i would check out that channel and then of course i actually have
done a few data camp tracks recently just to keep up to date with some of the uh the lm stuff i mean you guys have been a massive partner on my channel so i can't go without sharing that and yeah that's kind of where I'm at right now that's cool and I'm impressed that you've done some of the data camp tracks that's good for you I hope you learned something useful alright thank you so much for your time Tim thanks for having me let's do it again Thank you.
番組の概要欄(原文)
Landing an AI engineering job takes more than a good resume. Hiring teams see a lot of applications, and many portfolios never get opened because they're hard to run or hard to understand. Interviews range from recruiter screens to practical challenges, like building an API around a model, and some now allow AI tools as long as candidates can explain the results. What makes a portfolio project stand out? Do certifications help, and how many is too many? What does each interview round test? And how should you prepare for questions on RAG, fine-tuning, and caching? Tim Ruscica is the founder of Tech With Tim, a YouTube channel and brand teaching programming and AI skills, and of Kernel, a management company for tech and coding creators. He previously worked at Microsoft and ran DevLaunch, a program that helped developers land their first or next developer role. In the episode, Richie and Tim explore the different types of AI engineer roles, the software engineering skills employers still want, reviewing AI-generated code, building portfolios that get noticed, certifications, what AI engineering interviews look like, communication and soft skills, succeeding in your first months, and much more. Links Mentioned in the Show: DataCamp Certifications AI Plays Flappy Bird (NEAT) project Nate Herk's YouTube channel David Ondrej's YouTube channel Jack Roberts's YouTube channel Sentdex's YouTube channel Connect with Tim AI-Native Course: Intro to AI for Work Related Episode: How to Have a Machine Learning Career in 2026 with Marina Wyss, Senior Applied Scientist at Twitch New to DataCamp? Learn on the go using the DataCamp mobile app Empower your business with world-class data and AI skills with DataCamp for business
出典・この記事について
- 番組
- DataFramed(DataCamp)
- 元の回
- #387 How to Get Hired as an AI Engineer | Tim Ruscica, Founder of Tech With Tim
- 配信日
- 要約作成日
- 公式サイト
- www.datacamp.com/podcast
この要約・解説記事は、元の音声を文字起こしした内容をもとに AI(Claude)が日本語で作成したものです。 固有名詞や数値に誤りを含む場合があるため、正確な内容は元の音声でご確認ください。制作方法について
関連エピソード

営業・マーケ担当者はエージェントに関心がない:HubSpot CPTO ダンカン・レノックスが語る実用的なAIエージェント戦略
DataFramed

AIから「価値を生む仕組み」をつくる ― Deloitte USのAI責任者 Jim Rowan が語る企業AIの壁
DataFramed

クラウドの時代は終わったのか?ハイブリッド基盤とエージェント時代のデータガバナンス|Sergio Gago(Cloudera CTO)
DataFramed

チームとエージェントをまたぐ知能の協調法|Cisco Outshift SVP Vijoy Pandey
DataFramed

3か月でAI安全分野に転身する方法(キャリアアドバイザーMatt Beardによる記事朗読)
80,000 Hours Podcast

DataCampの学習者データが示す学び方のリアル | Pieter Moors(DataCamp リードデータサイエンティスト)
DataFramed