Principal Bioengineer Data ScientistFeb. 2025 – present
InsightRX — San Francisco, CA / remoteJan. 2020 – present
I lead pharmacometric and machine learning model development for a precision-dosing platform deployed across 1,000+ U.S. hospitals; five of my population PK models are live in the platform.
- Shaped an AI-native agentic workflow that runs clinical pharmacology analyses for pharma drug development (InsightRX Astra) as its domain expert: ran user testing, recommended UI and workflow fixes, and presented product demos to 5 pharma clients and 45 participants in an ACoP workshop
- Designed the GxP validation package for this agentic AI platform: human testing plus programmatic validation of nondeterministic outputs, intended to run each release
- Led scientific direction and a large-scale quasi-experimental study across 90,295 patients in 19 hospital systems, demonstrating that pharmacokinetic model selection meaningfully improves vancomycin AUC target attainment and reduces AKI rates in precision dosing contexts
- Developed the first population PK model to pool multiple aminoglycosides across the human lifespan, using real-world data from 5,659 patients across 130+ sites; now deployed in the platform
- Identified and corrected a systematic serum creatinine rounding bias in elderly patients that degraded renal function estimation and vancomycin/aminoglycoside dosing accuracy; the corrected model is now deployed in routine clinical care
- Developed and validated a novel vancomycin population PK model for critically ill patients on continuous renal replacement therapy using real-world data from 2,409 patients across 91 sites, outperforming five published models in predictive accuracy; now deployed in the platform
- Led ML development on a cross-institutional collaboration with Memorial Sloan Kettering — feature engineering, algorithm selection, and evaluation — producing a hybrid PK/PD-ML model that improved grade 3–4 neutropenia prediction precision from 47% to 61% across 9,121 patients
- Directed simulation-based analysis identifying precision dosing failures in a pharma client's conditioning regimen; presented recommendations directly to clinical decision-makers, resulting in a redesigned trial dosing strategy
- Built the analytics ETL pipeline behind real-time monitoring software supporting pharma drug development decisions, and presented the platform at ACoP
- Mentored a summer intern, now a senior engineer in medical devices
- Prototyped an LLM agent-based tool (GPT-4o) to assist pharmacometricians in NONMEM model development, then partnered with Pfizer to evaluate a more advanced version of this tool for pharmacometric model development against a manual workflow
- Led a tri-institutional collaboration with Lucile Packard Children's Hospital (Stanford) and Children's Hospital of Philadelphia to develop a pediatric voriconazole population PK model and optimized sampling strategies for model-informed precision dosing
Previously: Senior Bioengineer Data Scientist (2020–2025); Bioengineer Data Scientist (2020)