Software

Selected open-source projects. Most of my software is developed in the open on GitHub.

Numbl

Open-source MATLAB-compatible computing environment that runs .m files in the browser and on the command line; early stage (with Dan Fortunato).

numbl.org source

mip

Package manager for MATLAB and numbl: dependencies, precompiled MEX binaries, and path management; early stage (with Dan Fortunato).

mip.sh source

Neurosift

Browser-based visualization of NWB (Neurodata Without Borders) files and interactive exploration of the DANDI Archive, EMBER Archive, and OpenNeuro (with the CatalystNeuro team).

neurosift.app source paper

Figpack

Python package for interactive scientific visualizations as self-contained HTML bundles: portable, reproducible, and suitable for long-term archiving.

home source

Stan Playground

Browser-based editor and runtime for Stan models: edit, compile, run, and analyze results with no local installation (with Brian Ward and Jeff Soules).

live source

DANDI AI Notebooks

AI-powered chat assistant and automated notebook generator for exploring neurophysiology datasets in the DANDI Archive (with Ben Dichter, Ryan Ly, and Oliver Ruebel).

paper

Isosplit

Non-parametric clustering with no adjustable parameters, based on Hartigan's dip statistic and isotonic regression.

source

MountainSort

Spike sorting software built on isosplit clustering.

source

LINDI

Cloud-friendly file format and Python library for NWB data, compatible with HDF5 and Zarr and designed for linking to remote datasets (with the NWB team).

source

simple_ans

Simple, efficient lossless compression of integer datasets via Asymmetric Numeral Systems.

source

NWB Assistant

Experimental AI chat assistant for navigating NWB documentation.

live source

Stan Assistant

Experimental AI chat assistant for Stan, with access to the Stan User's Guide.

live source

Neurosift chat

Experimental AI chat assistant for exploring the DANDI Archive and OpenNeuro.

live

Spurious Discovery Tests

Synthetic experiments probing whether large language models avoid false discoveries when analyzing datasets with no underlying signal.

source

minicline

Command-line and Python interface for software engineering tasks with large language models.

source

MCMC Monitor

Real-time tracking and diagnostic visualization of Stan MCMC runs in the browser (with Jeff Soules).

source

nbfiddle

Web-based Jupyter notebook interface using browser storage, designed for self-contained analysis notebooks with remote data sources.

nbfiddle.app source

SpikeInterface

Unified Python framework for spike sorting and electrophysiology analysis; part of the original team.

docs source

FINUFFT

Fast, parallel library for the nonuniform fast Fourier transform on CPU and GPU; co-developed the original version with Alex Barnett.

docs source

Benchcompress

Benchmarking framework for compression algorithms on scientific data arrays.

results source

Remfile

File-like object for reading remote files over HTTP, optimized for h5py.

source

Also: the Concept Collection, a set of small proof-of-principle examples and interactive demos, many runnable directly in the browser.

Older projects

Figurl — shareable interactive figures; replaced by figpack
Kachery — content-addressed storage for sharing scientific data; replaced by figpack
SortingView — interactive spike-sorting visualizations; replaced by figpack
MountainSort 4 — replaced by MountainSort 5
SequenceTree — visual MRI pulse sequence programming