Latest from AI Hoodman
Short visual breakdowns of ML system design questions. Each post takes one real problem from ranking, recommendations or retrieval and explains the standard fix.
No. 013 · RetrievalApproximate Nearest Neighbor Search
How do you cluster a billion videos?
No. 012 · RetrievalRecommendation Funnel
How do you pick 10 videos from a billion in 100 ms?
No. 011 · LanguageSmart Compose
How do you make autocomplete fast enough?
No. 010 · RecommendationsPercentile Watch Time
Which video did they enjoy more?
No. 009 · RecommendationsInstagram Notifications
Which posts deserve a notification?
No. 008 · EmbeddingsUser ID Embeddings
How does a model tell User 1001 from User 1002?
Experience
- Jul 2024 to Present
TikTok
Machine Learning Engineer
- Sep 2023 to Jun 2024
Stanford Online Deliberation Platform
Senior Research Software Engineer
- Jun 2023 to Sep 2023
Microsoft Research
Research Software Engineer Intern
- Sep 2022 to Sep 2023
Stanford Reading & Dyslexia Research
Research Software Engineer
- Jan 2021 to Sep 2022
Amazon
Software Development Engineer
About
I am a machine learning engineer at TikTok in San Jose, where I work on ads: the targeting and user-modeling systems that decide who sees which ad, from gender models covering more than two billion monthly users to LLMs fine-tuned on viewing history. At that scale a one percent gain is a large number, and most of my work is finding the next one.
Retrieve fast and rough. Rank slow and precise.From AI Hoodman No. 012, Recommendation Funnel
Off the clock
Away from machine learning, I have been writing down my mom's recipes so they are never lost: biryani, dal, sheer khurma and everything in between.
Let's talk.
LinkedIn is the fastest way to reach me. Email works too.
How to reach me


