Research & Publications
Research That Starts From the Build Side
There is a gap between what academic AI research covers and what actually happens when you deploy it in production environments - whether that is a hospital, a learning platform, a consumer product, or an enterprise system. I am working to close that gap by formalizing what I have learned from real products, real failures, and real trade-offs.
Where I Am Coming From
Research Approach
What shapes how I think about this work.
Practitioner First, Researcher in Progress
I have spent several years building AI products across healthcare, EdTech, and enterprise technology - and I keep running into the same problem: the research literature does not cover what actually breaks in production. This page documents my attempt to change that by writing up what I have learned.
The Questions I Care About
How do you validate AI before it touches real users - whether patients, students, or consumers? How do you build evaluation frameworks for RAG systems that reflect real-world risk? How do you manage AI products when the model itself is a dependency you do not fully control? These are not questions I have fully answered yet - but they are the ones I am working on.
Research in Progress
Ideas I Am Developing
These are working ideas and drafts - not published papers. I am developing these based on patterns from my production work. None have been submitted for peer review yet.
Clinical Trial Patient Matching with LLMs
Documenting the approach we developed for LLM-assisted eligibility screening - what worked, what did not, and the accuracy/latency trade-offs we hit. Not a clean success story; more of an honest account of a messy problem.
Get Updates →Evaluating Enterprise RAG Systems in Healthcare
Nobody agrees on how to evaluate clinical RAG systems. I am trying to write up a methodology that is actually useful in production - covering factual accuracy, hallucination rates, and compliance considerations.
Get Updates →The Enterprise LLM Deployment Playbook
Patterns from running LLM deployments across healthcare, EdTech, and enterprise systems. The stuff that does not make it into model cards: infrastructure surprises, user adoption blockers, data quality issues that only show up in production.
Get Updates →Regulatory-Aware AI Product Development
A framework for building AI in regulated industries where compliance is part of the design from day one, not something you bolt on before launch. Drawing from experience navigating FDA, HIPAA, and enterprise security reviews.
Get Updates →Research Interests
Open Questions I Am Investigating
The questions I keep hitting that nobody has clean answers to yet.
Clinical AI Deployment & Validation
How do we move AI from research papers to clinical workflows? Interested in validation frameworks, clinician trust, and regulatory pathways for AI/ML medical devices.
Discuss →Enterprise GenAI Architecture
Production patterns for retrieval-augmented generation, LLM evaluation at scale, and the organizational structures that support enterprise AI adoption.
Discuss →AI Product Management Methodology
Developing new product management frameworks specifically designed for AI/ML products - accounting for model uncertainty, data dependencies, and iterative improvement cycles.
Discuss →Cross-Industry AI Transfer & Generalization
What patterns from healthcare AI deployment transfer to EdTech, CPG, and enterprise tech - and what requires rebuilding from scratch. The meta-question of how AI product knowledge compounds across industries.
Discuss →