To whom it may concern:

I would like to work on frontier AI models and products at your company.

I design, implement and test changes for Quora feed and digest recommendations. I know the system end-to-end enough that I can investigate any issue with the recommendation system. I run the ML systems design interview for recommendation systems and AI engineering.

I am the main maintainer for CLAUDE.md and hooks for the company monorepo. I developed the internal agentic query over the data lake to answer any data questions. I invented some coding interview problems that are AI-resistant for now. I still do LeetCode for fun.

I built the process to generate questions on recent events. I wrote prompts to label topics on content, and created the pipeline to backfill new topics.

I would like to highlight my personal project to fine-tune LLMs to solve ARC-AGI-2. My approach is to craft golden solutions and guide reinforcement learning with a modified version of GRPO.

I believe I have good taste in AI products, especially on how AI responses should be. I write about prompt engineering projects, instruction delivery, and reinforcement learning. I have also experimented with automated prompting, animating flash attention, profiling LLMs, AI competitive programming, and some Kaggle competitions.

I also write about things I believe will be built - prompt decaching, multichannel models, competitive programming supercommunity. I hope to help build any of these. Of course, I am happy to start out in any relevant position to prove my value.

I am currently based in the San Francisco Bay Area. I am happy to relocate to the City. I am currently on an H-1B visa. I am a Singapore citizen eligible for USA H-1B1 work authorization without lottery, unaffected by the 100k fee imposed on H-1B visas. I am eager to start working as soon as the visa process allows.

This cover letter is available on cover.huikang.dev. My resume is available on resume.huikang.dev.

I have worked on AI experiments during all my free time. I hope to work on AI full-time as well.

Thank you for reading.

Hui Kang