I build scientific AI products, MCP infrastructure, and research programs that turn complex evidence into decisions people can defend.
Public product figures verified July 30, 2026.
I design scientific AI systems around a clear chain from source to decision. The data model preserves provenance. The interface exposes uncertainty. The release process tests whether the system still performs as claimed.
This discipline drives Health AI's product intelligence and MCP infrastructure, as well as my research and teaching at Yeshiva University.
A model can produce an answer. A trustworthy system must also show why the answer deserves confidence.
A portfolio spanning health intelligence, MCP verification, regulatory evidence, and lifecycle validation.
A condition-aware intelligence system for ingredients and products. It preserves product identity, evidence, uncertainty, and health context across the checker, scanner, chat, API, and MCP.
MCP Queen probes MCP servers and grades the evidence behind each connection. It reports reachability, metadata quality, security checks, and reproducible receipts.
An MCP for source-linked FDA and CMS intelligence. Its 13 public tools support predicate research, device evidence review, and regulatory monitoring.
Visit Constat MCP →A 12-tool MCP that gives agents access to Clarity's product, ingredient, comparison, evidence, and condition-aware safety intelligence. Responses use explicit schemas and provenance.
Connect to Clarity MCP →A lifecycle validation method for high-stakes AI. It links requirements, implementation, governance, operational evidence, and runtime monitoring.
See the method →I connect evidence, product design, and operational controls around decisions that carry real consequences.
I designed the product strategy and shared data contracts behind Clarity's checker, scanner, chat, API, and MCP. The system includes ambiguity handling, citation controls, and release gates across more than four million scannable products.
Read the methodology →I reframed MCP discovery as a verification problem. MCP Queen tests server reachability, applies deterministic criteria, records evidence, and monitors change.
Inspect the registry →I built a field inspection workflow and scanner evaluated on 203 vehicles, then published a dual-VLM consensus study with an explicit cohort, method, and scope.
View the technical publication →
At Yeshiva University, I lead a research program that combines cross-seed reproducibility, interface energetics, metal-site plausibility, and benchmarking against experimental homologs and physics-based models.
Three tailored résumés, each organized for a distinct hiring context.
Technical success, deployment strategy, regulated workflows, adoption, and customer-facing execution.
Download PDF →0→1 product leadership, platform architecture, agent infrastructure, portfolio strategy, and delivery.
Download PDF →Evaluation, provenance, model behavior, evidence systems, computational chemistry, and research leadership.
Download PDF →Product and platform leadership, scientific AI deployment, evaluation and trust, advisory work, speaking, and research collaboration.