Healthcare workflow, software, and AI delivery
Independent consultant for healthcare operators and technical teams
Healthcare organizations lose time and operating capacity to fragmented data, manual review, brittle integrations, and uncertainty about where AI can create value. I work directly with operators and technical teams to identify the highest-value intervention, decide whether to build, buy, integrate, or automate, and carry the work through validation and handoff.
Ways to engage
- Focused assessment. A short observation of your workflow, ending in a concrete build, buy, integrate, or automate recommendation.
- Bounded implementation. A focused software, automation, data, or AI solution built, validated, and handed off with clear measures and ownership.
- Embedded delivery. Ongoing work alongside your operators and technical teams through a critical stage, owning delivery from decision to production.
Best fit
- Providers seeking measurable operational improvement.
- Healthcare organizations constrained by poor-fit or disconnected systems.
- Health systems and health-tech ventures needing technical delivery.
Selected experience
12 years building and leading production systems across early-stage startups, growth-stage healthcare companies, and Meta.
-
Zus Health
Built core patient-data infrastructure spanning normalization, deduplication, and entity resolution, unifying millions of records from thousands of healthcare sources.
-
Omada Health
Built the MLOps platform, standardizing model development, batch and real-time inference, evaluation, and production monitoring.
-
Meta
Automated model release for a neural-input wristband, cutting roughly two weeks of intensive engineering work to a few days of light supervision.
-
HealthLeap
Built a testing framework and stabilized core data and ML pipelines for reliable production integrations with health-system Epic environments.
Engineering capability
Good engineering judgment starts with choosing the intervention that fits the workflow, economics, and consequences of failure, whether integration, conventional software, deterministic automation, machine learning, or frontier AI. When AI is the right tool, I combine model judgment with deterministic state, business rules, human escalation, reproducible evaluations, and replaceable model and vendor boundaries.
I match the engineering approach to the stage and stakes, from a fast prototype that resolves the next uncertainty to a mission-critical system that must scale safely. Production systems are built around durable data models, explicit failure boundaries, traceability, security, controlled releases, meaningful telemetry, rollback, and clear ownership so they remain reliable, understandable, and operable as usage grows.
Availability
Available for focused assessments, bounded implementation projects, embedded engagements, or dedicated full-time consulting. Based in the Bay Area and willing to work onsite and travel regularly.