Daoyuan Li

Estimate, benchmark, and generate fine-tuning recipes for LLMs on consumer GPUs.

Retrieve + rerank over a closed label bank: LLM bi-encoders with self-mined hard negatives and a generative listwise reranker

One GPU. Full LLM workflow. Real benchmarks. No cloud required.

A lightweight toolkit for de novo molecular generation (SMILES/SELFIES; CharRNN, MolGPT, VAE)

Train and serve pairwise LLM judges (A/B/tie) with budget-aware multi-turn packing and position-bias correction