Experience¶
22 years of engineering and 18+ years leading AI research, data science, product, engineering, and platform work, 9+ of them at Director/VP level. In that time I've gone from hands-on research to directing distributed organizations, building complete AI capabilities from zero in cancer research, consumer AI, biotech, agriculture, clinical healthcare, and autonomy.
Autonomous Solutions, Inc. (ASI)¶
AI Infrastructure Engineering Manager¶
I lead AI infrastructure and the data platform behind autonomy and advanced prototyping at this robotics and autonomous-systems company. I grew a cross-functional team from 2 to 8: engineers across data, embedded, AI infrastructure, AI applications, and software, plus data scientists. Across IT, OT, and R&D, I'm speeding up the path from research to product while keeping deployments safe and operationally rigorous.
- Hybrid compute. Developed and implemented a hybrid strategy across on-prem, cloud, and GPU neocloud providers on a $1M+ annual budget. It serves AI, platform architecture, simulation, and other business units, with workloads from vehicle simulation to imitation learning.
- On-prem hardware. Directed hardware strategy and provisioning that connects vehicle blackbox data to imitation-learning and simulation workloads, making development pipelines 12× faster.
- Provisioning. Redesigned provisioning so most compute and IT requests take under 2 weeks instead of 6+ months, with infrastructure deployed as code (Terraform).
- Company-wide data platform and data science. Moved vehicle blackbox data into Azure and Databricks through a medallion architecture, enabling real-time visualization of vehicles in the field. Grew the platform from an AI R&D data solution into infrastructure serving every business unit and data type in a 500+ person organization, with data science and predictive modeling built on top.
- Latency. Led the team that took data from daily batch views to near real time (seconds) in Databricks and Azure Data Explorer.
- Research-to-product. Designed a leadership strategy that aligns IT, OT, and R&D so AI research reaches product quickly and safely.
Loom¶
Founder & Chief Architect¶
I designed, built, and operate Loom end to end: a production healthcare AI platform for precision medicine. More on the Research page.
- A knowledge graph of 2.3M+ patients with 2.3B+ nodes and 4.8B+ relationships, integrating 18+ biomedical data sources: 143M+ encounters, 135M+ clinical notes, 10.8M+ imaging studies, and 1.74B observations
- A 100K+ line Python codebase
- End-to-end MLOps: model registry and observatory, evaluation pipelines, drift detection, and vector databases for RAG
- Privacy-preserving federated learning across sites, using differential privacy and homomorphic encryption
- Home of the ADAPT-Q research program
GiftHealth¶
Director of AI Engineering¶
I led AI strategy and engineering at a PE-backed digital specialty pharmacy serving millions of patients nationwide, in a HIPAA-regulated environment. I owned the company's entire AI stack, co-owned AI governance and compliance, and worked directly with the CTO and executive team.
- Developed and launched an 18-month plan to take patient-facing products from no AI to AI-first, covering patient graphs, real-time adherence prediction, and agentic intervention
- Designed a graph-first, Neo4j-backed patient prescription knowledge graph and led the team into its build
- Led the agentic AI communications team and the vendor lift-and-shift of the voice and chat stacks, which handle high patient call volume
- Built the strategy for, and led delivery of, company-wide virtual support for the Medicare bridge program, spanning pharmacy logistics (demand forecasting, routing, inventory optimization) and patient customer-experience automation
- Grew the AI team from 3 to 8 engineers, data scientists, and prompt engineers, and worked with Product to pick the highest-value AI targets
- Led patient-facing AI strategy for pharmaceutical clients, including Lilly, and worked with patient-advocacy leadership to put guardrails on AI that exceed standards of care
Mayo Clinic¶
Creator & Team Leader, Agentic Knowledge Engine¶
I conceived, architected, funded, and led an enterprise agentic AI platform from idea to production. It serves 70,000+ users across clinical, business, and research operations. I built and led a core team of about 5, scaling to about 10 for projects.
- Production agentic and custom GraphRAG systems on a Neo4j knowledge-graph backbone: multi-agent orchestration, tool use, RAG pipelines, and real-time inference at sub-second latency (LangChain/LangGraph)
- Cut dataset preparation from about 6 months to under 1 hour
- Drove a ~1000× reduction in inference cost in the first month through architecture and model routing. Total infrastructure cost came in below a single junior FTE.
- Established evaluation, MLOps, and deployment standards, plus responsible-AI governance: bias, fairness, and explainability (SHAP, LIME, Fairlearn), model monitoring, audit trails, and HIPAA compliance
- Developed cognitive-complexity scoring that quantifies clinician workload, for workflow optimization and clinician wellbeing
- Partnered with Clinical, IT, Quality, Privacy, Security, and executive leadership
Technical Lead & Lead Solution Architect, Transformation Hub¶
- Led technical development of Transformation Hub products and core infrastructure, and acted as technical liaison to other Mayo teams
- Designed the Hub's technical and product architecture around innovation acceleration, knowledge generation, analytics and AI engines, and modular data and model assets
- Knowledge Hub: an LLM-built knowledge graph from scraped and aggregated data, exposed through APIs, chat, document generation, and GraphRAG integrations
- Intake & Triage: an AI web app that routes innovators to Center for Digital Health services and surfaces gaps, opportunities, and staffing needs
Senior Analytics Architect, Solution Enablement¶
- Led a team of analytics architects solving for the Center for Digital Health and enterprise needs
- Built an AI system for self-service analytics, dashboards, and insights across Mayo Clinic data
Clinical AI focus areas
The goal was to modernize predictive pipelines with knowledge graphs, agentic systems, and analytical AI, with a strong emphasis on clinical implementation:
- Multimodal models, from naive builds through fine-tuned foundation models and GraphRAG
- Predictive assessment that gives clinicians actionable early-warning signals
- Therapeutic planning using digital-twin simulation of treatment pathways
- Risk identification through early detection and scoring
- Cognitive complexity algorithms that quantify the load on clinicians in patient interactions
The aim throughout: reduce clinician fatigue, improve patient outcomes, fit into existing workflows, and stay transparent enough to earn trust.
Syngenta¶
Global Traits Digital Science Lead¶
I led trait (gene and phenotype) discovery and delivery in R&D and IT Digital Science across Syngenta's global organization. I set the AI/ML technical vision, strategy, and standards supporting tens of thousands of scientists in 50+ countries, and led a distributed team of 25+.
- Created and led a generative AI program to automate and optimize gene delivery, with an AI governance framework emphasizing trust, safety, bias detection, and explainability
- Researched and deployed foundation LLMs for protein design, and predicted construct expression levels as a researcher workbench
- Aggregated data platforms into knowledge graphs (dbt, Snowflake), and led a data-governance overhaul covering regulation, privacy, and government/FDA audit readiness
- Owned the full stack from data model to applications, including the APIs, MLOps, and DevOps that connect gene discovery to trait introgression
AgBiome¶
Head of Data Science, ML & AI¶
I built and scaled the data science, ML, and AI organization from 7 to 20 people (30+ at peak project demand) across data science, ML platform, data engineering, and analytics.
- Delivered 151 commercial products in 8 months through systematic engineering execution
- Built distributed ML pipelines over petabyte-scale data (Spark, Databricks, Delta Lake), and hand-built core platform components, including a custom feature store and model management
- Created GENESIS, a digital twin that combines genomic and geospatial modeling to speed up lead identification and customer delivery (press)
- Developed a 4-year data strategy adopted enterprise-wide; established MLOps, experiment tracking, model registry, CI/CD, and data-governance standards; navigated USDA/FDA compliance
- Brought generative AI and LLMs to genomic data, and built interpretable ensemble methods for mode-of-action discovery from small datasets
- Produced 2 publications and multiple patents
Firstleaf / Penrose Hill¶
Head of Data Science, ML, AI & Research¶
I built the data science and AI organization from 0 to 15 people, across data science, ML engineering, ads and growth, BI, and product analytics, distributed across the US and the Americas, and took it to IPO-ready maturity.
- Drove 100× user growth, from thousands to 300K+ members with >95% satisfaction, plus +40% customer lifetime value and +25% retention; a redesigned onboarding flow raised completion by 80%
- Built real-time recommendation and matching systems serving millions a day at 98%+ accuracy and sub-100 ms latency, with A/B testing, multi-armed bandits, continuous deployment, and a full platform (experiment tracking, model registry, monitoring)
- Created patented deep-learning algorithms on molecular and marketing data, and user profiles built from billions of data points per user
- Built self-serve experimentation tooling so product and marketing could run experiments without waiting on data science
- Developed data-driven product creation optimized with MCMC parameterization
- Recognized in Fast Company's Most Innovative Companies: Data Science (2022)
University of North Carolina¶
Postdoctoral Research Fellow, Precision Oncology AI¶
- Built statistical and ML models and large-scale data pipelines on HPC for precision oncology and epidemiology, including causal inference on observational clinical data
- Modeled noise in single-cell signaling during stress responses using live-cell imaging and ML on large datasets
- Built and maintained the lab's live-cell imaging infrastructure, and developed a suite of time-series algorithms from signal processing to ML
- Secured two fellowships, contributed to publications including Nature Communications, and mentored graduate students and postdocs
Thomas Jefferson University¶
PhD Graduate Research Fellow, Biochemistry & Molecular Biology (Computational Biology)¶
- Identified a novel cell-division checkpoint by combining biochemistry, live-cell imaging, and quantitative/computational analysis