I build ML and scientific software systems for molecular discovery: docking, molecular generation, ADMET/QSAR, structure prediction, chemical LLMs, spectra modeling, and agentic chemistry evaluation.

My work sits between research prototypes and usable infrastructure: models, benchmarks, reproducible pipelines, CLI tools, remote GPU/HPC workflows, and small product-grade tools that make molecular data easier to inspect and debug.

Projects · LinkedIn · ORCID · Kaggle

• AI-driven drug discovery systems — molecular docking, protein-ligand modeling, generative molecular design, ADMET/QSAR, retrosynthesis-aware evaluation, and virtual-screening workflows.
• Scientific evaluation infrastructure — benchmarks, run artifacts, leaderboards, report generation, pose validation, and reproducible model-comparison pipelines.
• Chemical LLM and agentic systems — evaluation ladders for reaction understanding, retrosynthesis, route planning, MS/MS tasks, tool-enabled reasoning, structured judging, and retry/validation guards.
• Structure and docking workflows — Matcha/GNINA/SMINA/Vina-style docking, pose filtering, ABCFold/Boltz/Chai/OpenFold/Protenix-style structure-prediction workflows, and remote GPU/HPC execution.
• Research tooling and UX — Python packages, CLI/TUI tools, macOS molecular previews, dashboards, automation, and developer workflows for computational chemistry teams.

• Cheminformatics and molecular modeling: RDKit, BioPython, Open Babel, GNINA, SMINA, AutoDock Vina, PoseBusters-style validation, symmetry-corrected RMSD, AiZynthFinder, OpenMM/GROMACS, Mol*.
• ML and deep learning: PyTorch, PyTorch Geometric, Transformers, molecular language models, graph ML, flow matching, reinforcement learning for molecular optimization, scikit-learn, XGBoost, CatBoost, Optuna.
• LLMs and agentic evaluation: OpenAI-compatible APIs, PydanticAI-style workflows, LLM-as-judge, structured outputs, tool registries, validation/retry guards, benchmark matrices, rubric-based evaluation.
• Scientific software and infrastructure: Python packaging, uv, Docker, micromamba, SLURM, Bash, GitHub Actions, pytest, Ruff, reproducible configs, JSON/JSONL artifacts, CLI/TUI design, remote Linux/GPU operations.
• Product/tooling: Swift/macOS Quick Look extensions, JavaScript/TypeScript, WebKit/Mol* integration, dashboards, automation, and scientific UX.

I am open to research and engineering collaborations in medicinal chemistry, molecular modeling, drug discovery, chemical LLMs, molecular spectra, and ML infrastructure for scientific work.