Yejun Yun
CS + Math at Yale, class of 2028
I'm interested in technical AI safety, currently working on risk-aversion and previously control. I founded Yale AI Alignment in 2024. I am a consequentialist and longtermist.
Research
Writing
Experience
-
Roblox, ML intern2026
Set up a distributed KV cache with Mooncake for Roblox's LLM assistant, cutting time to first token by 10%. Contributed fixes upstream to Mooncake (#3207, #3436).
-
Graph & Geometric Learning Lab, Yale2026
Research on memory-efficient long-context inference, using context distillation and token eviction.
-
Supervised Program for Alignment Research2025
Research fellow, mentored by Benjamin Arnav (OpenAI) and Pablo Bernabeu-Pérez (LASR Labs). This work became the ensemble monitoring paper. Third place in the Demo Day lightning talks.
-
Yale AI Alignment, founder and board member2024–now
A student group on AI safety with 300+ people in its Slack. Led weekly sessions on transformers, interpretability and safety techniques.