Giordon
Stark

Pure-Python implementation of the HistFactory statistical model for multi-bin histogram-based analysis. Supports fitting, limit-setting, and interval estimation with JAX backend for automatic differentiation and GPU acceleration. Enabled ATLAS to become the first LHC experiment to publish full statistical likelihoods. NumFOCUS affiliated project. Development also supported through IRIS-HEP Analysis Systems.

Python implementation of HS3 (HEP Statistics Serialization Standard), a serialization format for statistical models in high energy physics. Part of the DEMOS (Democratizing Models) effort, an initiative to broaden access to and reuse of statistical models from LHC experiments.

MadGraph-Pythia-Delphes reinterpretation pipeline that chains HEP simulation tools to constrain previously untested models of new physics using existing ATLAS search results. Configurable via TOML files with support for multiple containerization backends. MaPyDe allows one to run all of the various HEP toolings or chain them together and perform a quick analysis with the results, such as running CERN ATLAS SimpleAnalysis or pyhf.

Python package containing schemas and helper functions enabling analyzers to work with ATLAS datasets (Monte Carlo and Data) using coffea for columnar analysis.

Distributed histogram management service — fill remotely, retrieve snapshots, process large datasets across distributed systems. Developed a web dashboard as part of IRIS-HEP Analysis Systems work.

Python wrapper for the ATLAS ITk Production Database API, enabling programmatic access to database operations for the Inner Tracker upgrade.

Contributions to the ATLAS offline software framework: jet substructure calculations, Xbb tagger, jet reclustering, nanobind Python bindings for CP tools enabling columnar analysis. Led the migration of the entire offline software codebase from SVN to Git.

Documentation for USATLAS Analysis Facilities. Revamped the full documentation site — migrated from Read the Docs to GitHub Pages, improved the build and release cycle, modernized the contributor experience, and cleaned up outdated content.

Benchmarking framework for USATLAS Analysis Facilities, automating performance testing of compute nodes and storage across multiple facilities.

Custom GitHub Actions runner enabling CI/CD workflows to execute directly on UChicago Analysis Facility hardware, bridging GitHub Actions with HPC/HTC computing resources.

Skills marketplace for analysis users who deploy agents on the UChicago Analysis Facility, providing reusable Claude Code skills tailored to HEP analysis workflows.

Machine learning platform container images for JupyterLab and BinderHub deployments on the UChicago Analysis Facility. Provides pre-configured environments with HEP and ML software stacks for interactive analysis.

Analysis framework built on top of the ATLAS xAOD data model, providing configurable algorithms for event selection, object calibration, and systematic uncertainties. Used by multiple ATLAS physics groups (SM, SUSY, Exotics, Higgs, Trigger, Jet Calibration).

Python extension module for converting ROOT TTrees to NumPy arrays and vice versa. Widely used in the HEP community for bridging ROOT-based C++ analyses with the Python scientific ecosystem. Archived in 2021 with the community transition to uproot-based workflows.

Tool for scanning the phase-space of a physics analysis to identify performant variables for discriminating signal over background.

Matplotlib stylesheets conforming to the ATLAS experiment's official plotting style guidelines, enabling publication-quality figures without manual style configuration. Archived and contributed upstream into the scikit-hep/mplhep project.

MCP server for Rucio distributed data management — lets LLMs search datasets, check replicas, manage replication rules, and monitor quotas.

MCP server for ATLAS AMI (Metadata Interface) and the PMG cross-section database — lets LLMs find MC samples, retrieve cross-sections, and validate dataset classifications.

Python wrapper for CERN's Glance API, enabling programmatic access to publication data through both CLI and Python interfaces.

Python library for parsing protocol data from Dr. Storage dry air cabinets, extracting temperature and humidity sensor readings from binary data across multiple cabinet models. Used in the ATLAS ITk pixel module test bench at UCSC.

Monitoring and control software for the ATLAS gFEX trigger electronics system, providing real-time system statistics and environmental condition monitoring.

OpenEmbedded firmware layer for compiling Linux kernels from scratch for ATLAS Level-1 Calorimeter trigger instrumentation, used in the gFEX trigger electronics project.

Web application enabling small-budget and non-profit theaters to provide free captioning services. Keyboard-controlled navigation through caption content via websockets.

Online dictionary for Science, Technology, Engineering, and Math sign language videos to be stored, uploaded, and searched. Part of the CD2Bit accessibility initiative.

Maintainer of 30+ conda-forge feedstocks providing packaged builds of HEP tools and ITk software for the scientific Python ecosystem. Includes pyhf, mapyde, pyhs3, sherpa, hepmc3, fastjet, delphes, atlas-schema, itkdb, itksn, module-qc tools, yarr, klfitter, and many more HEP/ITk tools. Active in the HEP Packaging Coordination group (formed 2024), coordinating cross-experiment packaging efforts across the HEP community.

Co-administrator and co-owner of the Scikit-HEP organization (2017-present), a community effort to build a Python ecosystem for particle physics. Joined when root_numpy was brought under the Scikit-HEP umbrella; contributions span root_numpy, pyhf, and organizational governance.

Mathematica program for curve fitting and error propagation in the Caltech sophomore physics lab. Leveraged Mathematica 8's statistical modeling features to import variable-column data files, automate error propagation, and provide standardized output for fitting resonance peaks, damped oscillations, and similar lab measurements.

A tile-based 2D exploration game built to learn the then-new HTML5 canvas element. Features a 32×24 block map, coin collection, and physics-inspired player movement with configurable speed and rotation — all written in vanilla JavaScript and jQuery at a time when canvas support was barely in nightlies.

A jQuery plugin implementing a configurable typewriter animation effect. Handles HTML markup and special characters correctly, supports letter-by-letter, word-by-word, or phrase-based typing modes, and fires an onComplete callback on finishing. Uses jQuery's queue system so multiple calls chain naturally. Built as an exercise in jQuery plugin architecture.