Deploying Computer Vision for Surgical Instrument Tracking: What the Papers Don't Tell You
The gap between a published model and a production system in the OR is wider than you think.
The Agentic AI Stack: Architecture Patterns for Production Systems
Agentic AI in production is a systems engineering problem more than an AI problem. Here is the architecture that actually works — orchestration patterns, memory design, guardrails, and the observability layer you cannot skip.
State of AI in Healthcare 2026: Executive Summary
1,591 active AI clinical trials. 50 involving LLMs. 17 in Phase 3. Key findings from the State of AI in Healthcare 2026 report — with a CTA to the full members-only analysis.
RAG vs Fine-Tuning vs Agents: A 2026 Decision Framework (Not Just for Healthcare)
One of the most common questions I get from product managers entering AI for the first time is some version of: "We want to
What HIMSS 2026 Tells Us About Where Healthcare AI Is Really Going
HIMSS 2026 confirmed what many of us expected: the ambient scribe moment is over, agentic AI is the new battleground, and the validation debate has finally become mainstream. Here is my read on what actually matters.
LangChain vs LlamaIndex vs Haystack: The 2026 RAG Stack Comparison
Every production RAG system I have built or reviewed in the last two years has forced the same decision: LangChain, LlamaIndex, or Haystack? The answer
LLM Evaluation Framework: Beyond Benchmarks
Public benchmarks are misleading. Here's how to evaluate LLMs for your actual production use case — custom eval sets, automated scoring, and the cost-quality-latency tradeoff framework.
GenAI Build vs Buy: A Decision Matrix for Enterprise LLM Decisions
How AI Is Reshaping FinTech: From Fraud Detection to Autonomous Trading
Financial services is one of the most data-rich, highest-stakes industries in the world — which makes it both the ideal environment for AI and one of the hardest to deploy it in responsibly. Here's a practitioner's view of where AI is actually creating value in FinTech right now.
Multimodal AI Explained: Text + Images + Video in One Model
For most of AI's history, a model did one thing: text, or images, or speech. Multimodal AI does all of it simultaneously — and the product implications are bigger than most PMs realize.
How We Reduced Clinical Trial Matching Time by 73% with RAG
A real-world implementation of retrieval-augmented generation for matching patients to clinical trials at scale.
State of AI in Healthcare 2026: From Clinical Trials to Production
1,591 active AI clinical trials. 50 involving LLMs. 17 in Phase 3/4. A practitioner's deep dive into where healthcare AI actually stands in 2026 — the data, the regulatory landscape, what is not working, and predictions for 2027.
Mental Models for AI Product Decisions: When to Ship vs When to Wait
The most consequential decision I make as an AI PM is not which model to use or how to structure the evaluation pipeline. It is
LangChain vs. LlamaIndex for Healthcare RAG: A Technical Comparison
If you're building RAG (Retrieval-Augmented Generation) applications for healthcare - clinical decision support, medical literature search, protocol Q&A, pharmacovigilance - you&
