How I Can Help

I enjoy helping, especially on the topics of ML, RecSys, LLMs, and engineering. For a while, I responded to every email, and had an open calendar that anyone could schedule time on. However, that became unscalable. Thus, here are some guidelines on what I can—and cannot—help on. If you’re still unsure, shoot me a DM or email anyway.

• • •

First, I think it’s good to explicitly state what I’m not much help on:

• Content collaboration, even if it’s paid. If I’m made to write about something that I’m not naturally interested in, it’ll probably suck. >50% of calls with startups involve this: The answer is always “No”.

• How to get your first job in data/machine learning or how to get into big tech. There’s plenty of people sharing their experience online, which includes how they applied, prepared, interviewed, negotiated, etc.

• How to get into and do well in OMSCS. I’ve written an FAQ about it here.

• Referrals if I don’t know you personally. This approach will most likely not work and I rather not get your hopes up.

• Bugs, general DS/ML questions. Google and Stack Overflow will answer this faster.

• • •

Here’s what I think I can contribute meaningfully on:

• Investing in or advising your startup/team on data and machine learning topics, especially if applied in industry. This is my wheelhouse and I’m happy to help.

• How to structure a data team, frame an ML problem and design a system to solve it, working with LLMs, etc. Some of my thinking here.

• Reference checks on companies or managers if you’re considering a new job. I hear about data/ML teams in industry and may also be able to help via back channels.

• Introductions. Happy to provide intros to people I know, after checking with them that it’s okay. Do include a paragraph about yourself that I can send to them.

• Speaking at conferences or podcasts. Grateful for opportunities to share about building ML, RecSys, and LLM systems, building & leading data teams, etc.

If it falls into any of these buckets, send me an emailI read everything but receive too much to respond to all of it. or schedule some time.