Working on the Perception ML Foundation team to improve training efficiency and streamline MLOps for autonomous vehicles.

Mentored graduate students and postdoctoral researchers in the Chemistry department on machine learning researchCo-PI on award for over $20K in computational resources (XSEDE award CHE210034, PI Seonah Kim)

Proposed, planned, and executed a $1.8M dollar research effort across three institutions on deep learning for energy storage, resulting in high-impact publications in Nature Machine Intelligence and Patterns.Lead software architect on peer-reviewed software for reinforcement learning over action spaces defined on graphs (NREL/graph-env) (Python, Tensorflow, CI/CD)Lead NREL PI on a 3-year, $1M grant focused on predicting thermotolerance and pH tolerance of enzymatic transformations (NLP, torch)Project lead for a $275K/year subtask under the Co-Optimization of Fuels & Engines initiative. Developed explanatory model for particulate emissions for industry stakeholders using machine learning and Bayesian inference (scikit-learn, pymc).

Managed a million+ record SQL database with pipelines for ML model training. Published in Scientific Data.Trained a custom graph neural network (tensorflow 2 / keras), and deployed a containerized web application (Docker / FastAPI / Flask, bde.ml.nrel.gov) on AWS with over 8k unique users. Published in Nature Communications.Implemented CI/CD pipelines to automatically update, test, and deploy new models to website services.Pre-trained a BERT-style natural language processing model on 261M unlabeled protein sequences using over 1024 GPUs (torch), fine-tuned resulting model to improve protein annotations on limited data. Recipient of NREL's President's Award (2019)

Used metabolic modeling (python / cobrapy) to suggest unintuitive reaction knockout to improve yields of muconic acid from glucose. Improved yields in shake flasks from 7% to 41%, matching modeling predictions within a few percent. Published in Joule.Recipient of NREL's Director's Award (2017)