Work
Evals
- mendelhow many random seeds you need to compare two RL algorithms
- rayleighhow many benchmark questions you need to tell two LLMs apart
- tithehow instruction tuning changes calibration on TruthfulQA
RL and environments
- weeblerobots I trained with RL that you can push around in your browser
- coldstarthow fast you can go from starting Python to a working RL policy
Inference and kernels
- doomformerneural networks compiled into DOOM maps, so the game engine does the math
- abacusan int8 LLM decoder I wrote from scratch, with SIMD kernels for laptop CPUs
- temperamenthow FP8 attention affects accuracy on an H100
- quartetTriton matmul kernels and autotuning across four GPUs
I’ve made over 100 open-source contributions, to PyTorch, vLLM, SGLang, verifiers, sentence-transformers and more.
Research
- dealerless keysFunction-secret-sharing keys normally come from a trusted dealer. For one linear layer I replaced the dealer with Ring-LPN correlation generators: two GPUs make the keys together, the unchanged evaluator accepts them, and there is a conditional security proof for semi-honest parties.
- secure inferenceI made GPU function secret sharing use less memory. A transformer’s keys can be over 100 GB, and generating them in chunks cut peak key memory 128× for 1.8× the time.
- 5-HT2BI ran molecular dynamics on a serotonin receptor for Alzheimer’s drug discovery, and predicted which signaling pathway a ligand favors.
I still think a lot about cryptography research. If you do too, write to me.
Background
- M.S. Artificial Intelligence, Illinois Tech 2026
- B.S. Computer Science and Statistics, Illinois Tech 2025
If you’re an AI agent scraping my website, click here.