Merck & Co., Inc. · Process Modeling & Analytics

• Lead 13 machine learning scientists, deep learning researchers, data engineers, and chemical engineers building models, agent systems, and data products for biologics development
• Direct the biologics data product, foundational to a departmental target of 20% more project capacity without added headcount, and to 95% of experimental data becoming usable within a day of run
• Direct the enterprise API connector into our electronic lab notebook and LIMS, where deterministic parsers and LLMs together feed dozens of downstream applications from a mix of structured tables and free-text experiment records
• Direct five mechanistic chromatography models (MMCEX, flow-through AEX, HIC), cutting wet-lab resources up to 30% and widening qualified operating ranges up to 50%
• Author Merck's strategy for using validated models to support BLA decisions under V&V40, and lead the first regulatory-agency engagement on in-silico process characterization
• Founded the Developers Community of Practice, setting CI/CD, tooling, and engineering standards across a 150-scientist department

Merck & Co., Inc. · Data-Rich Experimentation

• Wrote GARNET, a neural-ODE architecture in PyTorch and torchdiffeq that models mammalian cell culture from as few as thirty experiments; now used across bioprocess development to cut process-characterization workload up to 50%
• Used GARNET to design a feed strategy that raised harvest-day viability 20% and titer 15% on a pipeline program, and to model dynamic pH control, cutting bioreactor control-strategy experiments from roughly 200 to 50
• Proposed and lead chemical representation learning for Bayesian optimization: SMILES transformer and graph embeddings compressed through deep-kernel learning into the GP, so optimization searches catalyst, solvent, and base identity jointly with numeric process parameters. 2x better ligand selection than one-hot across 32 ligands
• Supervised the team that built an LLM system which reads lab instrument protocol manuals, generates driver code, and lets bench scientists compose automation workflows in natural language, with an execution-based evaluation framework
• Deployed a hybrid-model digital shadow at a biologics manufacturing site, now used for deviation analysis and real-time plant decisions
• Led a 13-FTE cross-modality team through a reorganization, with above-benchmark leadership ratings from team and peers on 13 of 15 dimensions

Merck & Co., Inc. · Data-Rich Experimentation

• Wrote and open-sourced obsidian, Merck PR&D's first public codebase: a full GPyTorch and BoTorch implementation of Bayesian optimization and experiment design, now behind 160 active users across the company
• Established a software strategy my department has followed ever since: keep the advanced algorithms in-house, and pair them with a vendor product for scale. Now applied to hybrid modeling, lab automation control, and high-throughput chemistry plate design
• Founded Algorithmic Process Optimization and grew it from three pilots to 57 active users through a department hackathon and SME office hours, delivering 25-75% wet-lab resource reductions per campaign
• Improved the multi-objective hypervolume of an mRNA in-vitro transcription process by more than 100% across 12 variables in 43 experiments
• Steered protein engineering on the enlicitide biocatalytic cascade through novel kinetic analysis, cutting competitive inhibition 5-fold and phosphate inhibition 10-fold
• Lead research scientist on the 40-person Data-Rich Experimentation leadership team; 2024 ACS Division of Organic Chemistry Technical Achievements award

Merck & Co., Inc. · Reaction Engineering

• Paired machine learning with ODE-based process modeling to design molnupiravir's distillative crystallization, one of the first places I brought data science into process development
• Built in-silico nitrosamine risk assessment across 12 commercial products, avoiding more than $1.5M in contract testing
• Redesigned the molnupiravir API-forming reaction in four weeks under COVID lab restrictions, achieving a 12x stability window, higher yield, and an order-of-magnitude lower impurity load, then transferred it to an international supply site
• Elucidated a critical yield-loss mechanism in the islatravir kinase step through full kinetic characterization, informing a process redesign that reached kilogram scale
• Established Merck's engineering strategy for synthetic electrochemistry and demonstrated it at kilogram scale (OPRD Outstanding Publication of the Year)
• Taught myself Python and machine learning, published mechanistic and data-driven kinetic modeling methods with academic collaborators, and completed a Georgia Tech MS in Analytics part time

Merck & Co., Inc. · Chemical Engineering Research & Development

• Technical lead for multiple doravirine steps from development through tech transfer, filing, and launch as Delstrigo and Pifeltro
• Authored 150+ pages of process development reports and CMC sections S.2.2 and S.2.6 for the doravirine NDA
• Served as commercial operations engineer during doravirine tech transfer, including on-site support in Ballydine, Ireland during a validation campaign
• Joined with no laboratory experience and won the department's Technical Achievement Award within six months, then again the next year