Selected Work

Selected work

A small optimization demo, followed by examples from process development, modeling, electrochemistry, and biologics modeling.

↳ Try it yourself

Optimize a process you can't see.

A small two-parameter optimization demo. Click points on the surface, or let the optimizer suggest the next experiment.

0experiments

·best so far

Click the surface to run an experiment.

Reveal the true surface

The optimizer uses a Gaussian process surrogate with Expected Improvement, similar to obsidian, and runs in the browser.

01Process Development

A COVID-19 antiviral needed a process redesign on a compressed timeline.

Challenge: Develop a robust API-forming reaction and control strategy in four weeks.
Approach: Ran hundreds of high-throughput reactions, built kinetic models, and used simulation analysis to support a complete process redesign.
Result: More than a 5x wider stability window, an order-of-magnitude lower impurity burden, higher yield, and improved robustness for the manufacturing route.

02Electrochemistry

We took an electrochemical oxidation from grams to kilograms and wrote down the workflow so other teams could repeat it.

Challenge: Take a clean, selective electrochemical oxidation from grams to kilograms without losing control of it.
Approach: Built a workflow for scaling electrosynthesis, ran a kilo-scale thioether-to-sulfone oxidation, and engineered a spinning three-dimensional anode for a multigram Hofmann rearrangement.
Result: A kilogram-scale electrochemical oxidation, named OPRD's 2022 Outstanding Publication of the Year.

03AI & Modeling

I set up the tools and training so process chemists and engineers could run Bayesian optimization on their own experiments.

Challenge: Adapt Bayesian optimization for day-to-day process-development use.
Approach: Founded Algorithmic Process Optimization, built and taught the tooling, and published obsidian, the department's first open-source codebase.
Result: More than 100 scientists onboarded across small-molecule and biologics programs; individual campaigns met optimization targets using 25 to 75% fewer wet-lab resources.

04AI & Modeling · Biologics

GARNET moved from complex small-molecule reaction mechanisms into cell-culture and fermentation models for upstream development.

Challenge: Model a fed-batch cell culture, with about 15 metabolite and in-process measurements and 20 process parameters, from as few as 30 experiments.
Approach: Wrote GARNET in PyTorch: neural networks inside an ODE model, built to handle bolus feeds without restarting the integration.
Results: Improved titer 15% and harvest-day viability 20% in a lab proof of concept on a separate pipeline program. Modeling pH control let the team choose a right-first-time characterization approach with under 40 experiments, against more than 100 on a benchmark program.
Deployment: Also deployed in a hybrid-model digital shadow for deviation analysis, troubleshooting, and real-time plant decisions at a biologics manufacturing site.

The papers behind these projects are on the research page.