Research code is only useful if others can run it. 48 repositories across 8+ organizations, from a 128,000-atom eigensolver to an LLM agent framework. Everything below is public on GitHub.
Agentic framework connecting natural-language requests to molecular construction, simulation (ASE, MACE, RDKit, and external QC codes), analysis, and reporting through LangGraph tools, a CLI, a Streamlit UI, and an MCP server; the software behind the Communications Chemistry 2026 paper and the ChemGraph-XANES and multi-agent screening extensions.
Quantum Thermochemistry Calculator: integrates Open Babel with MOPAC/NWChem/Gaussian/Molpro calculations and MESS partition functions to generate NASA polynomials automatically; basis of the 2019 butane-oxidation thermochemistry paper.
Python framework for running quantum-chemistry workflows (single-point energies, geometry optimization, frequencies, thermochemistry) at scale with QC codes and machine-learned potentials (MACE, FAIRChem UMA), writing Parquet datasets for downstream analysis.
Agentified evaluation framework, built on A2A and MCP, for benchmarking code-generation agents that write PETSc programs.
Gaussian approximation potentials plus training/validation data for graphene, silicene, h-BN, h-AlN, and h-GaN, accompanying the Nanoscale 2023 paper.
Gradio/Hugging Face leaderboard for an agentic-AI benchmark in computational chemistry, ranking LLM agents on ChemGraph-style tasks.
Streamlit + DuckDB dashboard (live at iqc-dashboard.streamlit.app) for browsing, filtering, and 3D-visualizing large IQC result sets molecule by molecule.
Pure-NumPy library for deterministic, rotation/translation/permutation-invariant 3D conformer identifiers of the form `<Hill formula><state tag>-<geometry hash>`, with a draft specification and accompanying paper.
Documentation and drivers for the Shift-and-Invert Parallel spectral transformations (SIPs) eigensolver in SLEPc, demonstrated on 128k-atom DFTB systems (J. Comput. Chem. 2016).
ALCF AI-for-science training series materials; contributions to lecture/tutorial content.
High-throughput workflow and job automation for HPC; small contributions.
Interactive, simplified HPC from Jupyter notebooks (job submission, monitoring, and visualization helpers).
Jupyter notebooks used by the ALCF data-science group, plus the `ganymede` notebook collection.
Exploration of Docker/Shifter/Singularity containers for HPC with MPI benchmarks; companion recipes in `container`, `container_nwchem`, and `mopac_container`.
Notebooks for the ALCF JupyterHub service.
Multiresolution Adaptive Numerical Environment for Scientific Simulation; contributions to build/CI and integration work.
Top-level user-facing NWChemEx repository; plus contributions to TensorWrapper, NWXCMake, Utilities, ChemCache, Integrals, and CI configuration.
Inversion-of-control plugin framework for modular exascale scientific software (J. Chem. Phys. 2023).
Massively parallel variational subspace valence bond (VSVB) code (J. Comput. Chem. 2019).
C++ library describing chemical systems (molecules, basis sets, wave functions) for NWChemEx.
Benchmarks of distributed eigensolvers (Elemental GHEP and others).
Integration of the SLEPc SIPs eigensolver into SIESTA (J. Comput. Chem. 2018).
Documentation and drivers for the Shift-and-Invert Parallel spectral transformations (SIPs) eigensolver in SLEPc, demonstrated on 128k-atom DFTB systems (J. Comput. Chem. 2016).
PSCF: parameterized self-consistent-field code (MINDO/3, DFTB-style) for large sparse systems built on petsc4py/slepc4py; testbed for the SIPs spectrum-slicing eigensolver.
Parallel runtime abstraction layer (MPI, hardware resources) for NWChemEx.
Simulation development environment / SDK defining the property types and module APIs of NWChemEx.
C++ benchmark suite comparing eigensolver libraries on ALCF systems.
Python package computing matrix inertia (eigenvalue counts) with PETSc and SLEPc, published on PyPI.
Fortran code for many-body vibrational structure calculations (VSCF/VMP/VCC-type methods).
Log files, parsers, and plots for QETSc/SIPs scaling runs.
Agentic framework connecting natural-language requests to molecular construction, simulation (ASE, MACE, RDKit, and external QC codes), analysis, and reporting through LangGraph tools, a CLI, a Streamlit UI, and an MCP server; the software behind the Communications Chemistry 2026 paper and the ChemGraph-XANES and multi-agent screening extensions.
HPC workflow that couples generative AI with simulation to design new metal–organic frameworks at scale.
Agentified evaluation framework, built on A2A and MCP, for benchmarking code-generation agents that write PETSc programs.
CrewAI multi-agent system that turns computational-science papers into runnable code, with self-testing and iterative improvement.
Gradio/Hugging Face leaderboard for an agentic-AI benchmark in computational chemistry, ranking LLM agents on ChemGraph-style tasks.
LangGraph agent that gives a natural-language interface to the PubChem database with pluggable LLM providers (OpenAI, Gemini, Claude).
ALCF systems user documentation (Aurora, Polaris, AI testbeds); one of the most heavily edited repositories in this history and the knowledge base behind AskHPC/HPCBot.
Command-line interface to Argonne's Argo LLM API with modular system prompts, predefined tasks, file ingestion, and logging.
Dr. Doc: LLM tool that fixes grammar, formatting, and broken links in documentation using the Argo API.
RAG chatbot and QA-generation pipeline over HPC documentation (Chroma vector store, OpenAI/Argo models); precursor of the AskHPC paper (SC Workshops 2025).
pip-installable CLI to upload files to GitHub repositories without cloning (wildcards, target directories, UUID names).
React/TypeScript/Supabase web directory of AI tools for science, grown out of the Intro to HPC Bootcamp list, with visitor ratings.
Parses `.ninja_log` build logs and reports the slowest build targets.
Master Equation System Solver for pressure- and temperature-dependent phenomenological rate constants; recent contributions to build system and packaging.
Personal development copy of the Quantum Thermochemistry Calculator (Python modules for quantum thermochemistry calculations).
Quantum Thermochemistry Calculator: integrates Open Babel with MOPAC/NWChem/Gaussian/Molpro calculations and MESS partition functions to generate NASA polynomials automatically; basis of the 2019 butane-oxidation thermochemistry paper.
Automated workflow driver for theoretical kinetics data in the AutoMech suite.
Early reaction-class driver for automated kinetics calculations in the PACChem ecosystem.
Python framework for running quantum-chemistry workflows (single-point energies, geometry optimization, frequencies, thermochemistry) at scale with QC codes and machine-learned potentials (MACE, FAIRChem UMA), writing Parquet datasets for downstream analysis.
Streamlit + DuckDB dashboard (live at iqc-dashboard.streamlit.app) for browsing, filtering, and 3D-visualizing large IQC result sets molecule by molecule.
Pure-NumPy library for deterministic, rotation/translation/permutation-invariant 3D conformer identifiers of the form `<Hill formula><state tag>-<geometry hash>`, with a draft specification and accompanying paper.
Project materials for 'Evaluating Large Language Models for HPC Education' in the DOE Intro to HPC Bootcamp (undergraduates use LLMs to generate and evaluate HPC teaching content).
Jupyter tutorials (ML, LLMs, web scraping, fine-tuning) for the AI for Energy Justice project, whose goal is a question-answering model on energy-justice reports and legislation.
LLM-based automated scoring of renewable-energy laws against the Initiative for Energy Justice scorecard, run on HPC.
Web app that lets kids design a bracelet encoding text in binary — an outreach activity.
ChemGraph, ALCF user guides, HPC assistants
IQC, HashMol3D, chemistry agents
exascale quantum chemistry software
automated thermochemistry (QTC)
automated kinetics suite
GAP potentials for 2D materials
PETSc agent benchmark
valence bond code