I build scientific instruments for data
Fifteen years of helping researchers, product leaders and organizations work directly with their data
The dots behind this page are my tweets since 2008, highlighting in sync with each roll. Scroll to the end to explore the map.
Now
- Designing and building an agentic platform for scientists and engineers.
- Applying frontier AI to domain specific workflows.
- Developing moonshine an open-source tool for writing interactive explanations together with coding agents, in the spirit of Distill.
Previous
- Latent Scope , an open-source tool that embeds, maps, clusters and labels unstructured text so you can browse it as a map.
- A zoomable map of the human genome laid out on a Hilbert curve so all 3 billion base pairs stay browsable at every scale.
- Trained parse autoencoders for sentence embeddings available as a browsable taxonomy of their features.
- Data interfaces for research teams, including Altius, Topology, ARC Prize and Stability AI.
- Organized the Hidden States unconference in San Francisco.
- Developed features and prototypes for simplifying complex data workflows.
- Taught data visualization to thousands of data practitioners.
- Collected and mapped data in public: wildfires, air quality, COVID, elections and emoji.
- Co-authored four articles on Distill: How to Use t-SNE Effectively, Four Experiments in Handwriting with a Neural Network, The Building Blocks of Interpretability and Activation Atlas.
- Helped the Creative Lab open source 50 million drawings as the Quick, Draw! dataset, and worked with Google Brain on the sketch-rnn demo.
- Gave Machine Learning for Visualization at OpenVisConf 2018 (video). It is still the project that motivates the rest.
- Built data tools for the 300-person Google Cloud design team. In the largest one we treated user sessions as sequences, trained an RNN on them, then clustered and mapped the results to learn how people use different products together. It was the sketch-rnn idea pointed at product usage.
- First hire and founding software engineer.
- Designed and built the original data infrastructure, the reporting features used by the largest customers, and the internal query tools.
- Evolved versions of those were still in use when Lever was acquired for $500m in 2022.
The d3.js community
Organizer
2012 to 2019
- Built Tributary, a live-coding environment for d3.js, and then Blockbuilder, an in-browser editor funded on Kickstarter, with search over 40,000 community examples.
- Co-founded the Bay Area d3.js meetup in 2012, which grew past 5,000 members, and co-organized d3.unconf for five years.
- Wrote The Hitchhiker's Guide to d3.js, The Trouble with D3, How do you learn d3.js? and Touring the D3 Ecosystem.
- Hundreds of small experiments as blocks, notebooks and videos.
Where it started
Florida State University and before
Student
2003 to 2012
- M.A. in Scientific Computing at FSU. My thesis put a GPU fluid simulation (smoothed particle hydrodynamics ) inside the Blender Game Engine, and I wrote the Adventures in OpenCL tutorials along the way.
- Summer internship at Lawrence Berkeley National Lab
- GIS web programming at FREAC, building open-source versions of Esri maps with PostGIS and OpenLayers.
- Molecular Expressions is where I first learned to program visualizations: interactive tutorials about how microscopes work, after a high school spent taking pictures through them.