JAY ZENITH
I build post-training and evaluation systems for tool-using agents: data, environments, verifiable rewards, and RL training infrastructure.
- glyph
End-to-end post-training for a 4B Rust coding agent, built on a shared ChatML tool runtime across SFT, verifier RL, and evaluation. [blog] [environment] [raw evals]
- llama.cpp - CUDA upstream
Added GGML_OP_FILL on the CUDA backend for the Qwen3-Next path; removed a CPU fallback. Merged upstream.
- llama.cpp - sampling hot path
Cut a redundant O(vocab) allocation from token sampling. 1.9-2.2x on the sampling microbench.
- mini-sglang - Mistral support
Sliding-window attention for Mistral-7B, validated against Hugging Face past 6k tokens.
- pd disaggregation benchmark
A/B harness for prefill/decode split vs colocated serving.