I am currently the Principal Artificial Intelligence and Machine Learning (AI/ML) scientist at Form Bio, a spin-off of Colossal Biosciences, where I am working on bioinformatics problems related to genomic editing, therapeutic development and conservation biology. Most of that work now runs through large language and genomic foundation models, including agentic pipelines that plan and execute analyses over public sequence archives, large sequence to function models fine-tuned and evaluated for variant and regulatory-element prediction, and rubric-driven scoring systems for therapeutic candidate design.I previously worked at HRL laboratories in the Center for Human Machine Collaboration where I researched methods for improving applied machine learning and autonomous systems. I got my PhD from the Computational Cognitive Neuroscience Lab at the University of Colorado Boulder with Randy O’Reilly. I spent a few years between undergrad and grad school at New York University with Lila Davachi using fMRI to explore the neurobiology of episodic memory. I did my undergraduate at the University of Minnesota in Minneapolis where I studied Physics, and got involved in computational modeling with Chad Marsolek.

With each move into a new domain (cognition, autonomous agents, genomics, and orchestrating LLM systems) my curiosity and perspective broaden while my technical skills and scientific instincts sharpen. The more I can learn in each shift, the better.

Find out more through my publications, CV, or more generally in my resume.

Research Interests

I’ve found success in the role of an applied scientist developing machine learning solutions to advanced research problems. My interests include understanding and developing intelligent systems (human and artificial); analysis and visualization of complex, high-dimensional data; and quantitative approaches to art, music and aesthetics. I have extensive experience in academic and industrial approaches to research, product development and human studies.

I started my research career working to understand different learning and memory systems and how they can be related to underlying neurobiology. This got me interested in several research domains including psychology, neuroscience, and cognitive science. My interest and expertise in neural network systems preceded the modern AI era, which opened many doors to new domains including autonomous systems, generative neural networks, genomic editing, viral vector therapeutics, and conservation biology.

I’ve previously applied these research interests to the domain of lifelong machine learning under the DARPA L2M program. Here autonomous systems are expected to continue to accumulate knowledge beyond a single dataset or task. My work focused on using neural network models to generate samples of previous experiences to preserve a reinforcement learning system’s performance while also allowing it to continue to learn new tasks.

My current work is at the intersection of large language models and biological sequence data. On the sequence side there are large genomic and protein foundation models: Evo, ESM, and Enformer/Borzoi-class regulatory models applied to cross-species genome engineering, and viral packaging of transgene DNA. On the language side it means treating the LLMs as part of the research tooling rather than a text generator. I’ve built agent pipelines that query SRA, GEO, and ENCODE, assemble evidence, and score candidate designs against explicit, versioned rubrics. The most interesting part of shared between them: what the models have actually learned, how to interrogate its outputs to improve confidence and consistency, and how do you evaluate a system when the standard hold-out sets are insufficient.