Open-source tools from the group. Everything is on GitHub under jkitchin.
1,264 GitHub stars across these packages. Numbers update live from GitHub, PyPI and crates.io.
The scimax scientific computing environment for VS Code. Org-mode and Markdown editing, runnable source blocks with Jupyter kernels, org-ref citations, a research journal and capture templates, Marp slide decks, and export to PDF, Word and Jupyter notebooks.
A mixed-integer nonlinear programming modeling language and solver, using spatial branch and bound over convex relaxations with interior point subsolvers.
A pure-Rust family of interior point solvers that began as a port of Ipopt, now covering nonlinear, conic and quadratic programs.
A pure-Rust sparse symmetric indefinite linear solver with certified inertia counts, built for interior point optimization.
Deterministic global optimization of plain JAX functions, using McCormick relaxations of the jaxpr and branch and bound for a certified lower bound.
Extensions that add domain models and methods to the discopt modeling language.
Design of experiments for discopt, from active-learning optimization and classical factorial and mixture designs to model-based D-, A- and E-optimal design.
Microkinetic modeling for discopt, with catalysis kinetics, degree of rate control, mechanism discovery and selection, and electrochemistry.
Variable aggregation, a reduced-space presolve for nonlinear optimization, following Naik et al. (arXiv:2502.13869).
Application builders for AC optimal power flow and pooling problems in discopt.
An optimization course and a tutor command-line tool built on discopt.
A JAX framework for differentiable chemical process flowsheets, for simulation, optimization and technoeconomic analysis.
Differentiable soft decision trees in JAX with built-in uncertainty quantification, scikit-learn estimators and MILP export.
Linear and hyperplane decision trees in pure JAX for piecewise-linear surrogate models.
Iterative Bayesian additive regression trees in JAX, which find compact symbolic descriptors with uncertainty and export them to discopt for global optimization.
A Tennessee Eastman Process simulator in Rust, with Python wheels and a browser app.
A symbolic regression engine built on JAX.
Python computations in science and engineering. A book and a library of helpers for numerical methods, regression and more.
An ASE-compliant Python interface to VASP.
Modeling materials using density functional theory with VASP and the Atomic Simulation Environment.
Curate and search your scientific literature with natural-language queries over a local database of papers.
A personal knowledge base where an LLM distills papers, notebooks and data into an interlinked org-mode wiki.
Install and manage AI agent skills, subagents, commands, hooks and MCP servers from any git repo, for Claude Code, Codex, OpenCode, Gemini and more.
Superseded packages, kept here for reference.
A memory-safe interior point optimizer for nonlinear programming, written in Rust. Superseded by pounce.