1

How to cut into AI Compiler fields?

Profile picture
Mid-Level Software Engineer at Taro Community9 months ago

I have 2 years of experience in big tech product related teams (Typescript frontend and C# backend) and I want to pivot to more low level teams for better on call and team cultures. AI compiler / accelerators seem like a growing field. What are some good resources to learn?

I checked out one Microsoft Maia team that have 12 people. 10 of them are principal level. The team looks highly specialized, which makes cutting into it even harder.

80
2

Discussion

(2 comments)
  • 1
    Profile picture
    Tech Lead/Manager at Meta, Pinterest, Kosei
    9 months ago

    Feels hard to get a job in that exact role -- can you find something that is directionally correct? Since you work on TS frontend, moving closer to infra or language development will give you a better sense of what life would be like in AI compilers.

    This discussion is very relevant: https://www.jointaro.com/question/eFr9vsyTSiYLCbtsj7RR/whats-the-most-effective-way-to-switch-domains-in-tech/. In particular, Cat's answer:

    Feel extremely comfortable failing and losing face. That's a key trait to learn quickly in a domain you don't know much about. Do that as early as possible.

  • 0
    Profile picture
    Tech Lead @ Robinhood, Meta, Course Hero
    9 months ago

    This seems like a very narrow field. I Googled for "AI compiler resources" and most of the results are just general AI stuff. I only found this Reddit thread: https://www.reddit.com/r/Compilers/comments/14z6gwd/any_recommandation_for_studying_ai_compilers/

    That being said, side learning is almost never meaningful when it comes to domain switches. Very few hiring managers are going to believe the tutorials, certificates, and courses you cobbled together on the side are enough for you to come in and pivot into this domain to start pushing code at a polished production-level. If an online resource has a lot of traction, it has become a commodity and is too easy to achieve.

    The real way to pivot is to establish yourself as a high-performer, show a mastery of the meta-skill of quick learning (i.e. learning how to learn), and build up relationships, particularly with your manager.

    From this post, I'm guessing that you work at Microsoft. If you do and are a high-performer with a good relationship with your manager, talk to your manager about an internal transfer so they'll put in a great word for you. That's how you'll convince the Microsoft Maia manager to bring you on (if they have headcount), not by going through some online resources in your free time.

    Something else you can try is reaching out to the engineers on the Maia team and offering them a free ☕, 🍵, or 🧋 for some of their time. Ask them about the field, and if they like you enough, they might volunteer to mentor you. Follow the advice here: [Masterclass] How To Build Deep Relationships Quickly In Tech

    I know several engineers who switched from SWE into MLE at Meta (so a similar switch to what you're describing) successfully, and this is the playbook they all followed. I talk about this more in-depth here: "How to transition from back-end development to distributed systems?"