Agentic AI · Pharmacy & Health-Tech
I'm Paul Doolittle, a retired pharmacist with 32 years of practice. Through Data Compounded, I build MCP servers and clinical decision support tools that give AI systems safe, structured access to drug data, EHRs and PBMs.
MCP (Model Context Protocol) is the open standard that lets an AI model use a tool or data source without custom plumbing. It's how an assistant can look up a real drug label instead of guessing.
Interaction checking, dosing logic, and formulary rules encoded the way a pharmacist actually reasons through them — not a generic API wrapper.
Purpose-built Model Context Protocol servers that give agentic AI systems safe, structured access to drug data, EHRs, and PBMs.
Secured, monitored, production infrastructure — HTTPS, reverse proxy, process supervision — so the prototype is the product.
I believe MCP will do the same for AI, doing for tools and data what HTTP did for documents: connecting any model to any tool, without anyone having to think about how. That's why I build servers.
Structured FDA drug labeling and adverse event data for agentic systems — two read-only tools, scoped by what the data can and can't tell you.
Read the case study →My first production deployment: a small MCP server for character and word counts, live at mcp.datacompounded.com.
Thirty-two years behind the pharmacy counter teaches you where clinical systems actually break — the interaction the software missed, the formulary rule nobody encoded, the handoff that lost context.
Today I bring that judgment to agentic AI. After 15 years of retirement, I found I missed having a hard problem in front of me, and that led me to AI engineering. Through Data Compounded, LLC, I build MCP servers and clinical decision support tools that give AI systems safe, structured access to drug data, EHRs and PBMs.
I also owned and operated a relief-pharmacist staffing company for 11 years, placing pharmacists across more than 40 pharmacies. I've seen the profession from both sides: the day-to-day of dispensing and counseling, and the business of keeping pharmacies staffed and running.
I work the way I practiced pharmacy: carefully, and checked twice. I take on a small number of projects at a time so each one gets full attention, and I don't call anything done until it's deployed, tested and verified. My OpenFDA MCP Server is live and documented, and that's the standard every project gets.
| Engagement | Best for | What you get |
|---|---|---|
| Clinical MCP server build | Teams needing drug data, interaction, or formulary tools wired into an AI workflow | A deployed, documented server your AI workflow can call |
| EHR / PBM connector | Health-tech products that need structured, agent-readable access to existing systems | Agent-ready access to the systems you already run |
| Deployment & infrastructure | Projects that have a working prototype but need it live, secured, and monitored | Your prototype, live, secured, and monitored |
| Advisory / architecture review | Teams validating clinical logic or MCP design before they build | A clear review of your clinical logic and MCP design |
I build MCP servers and clinical decision support tools for pharmacy and health-tech teams, drawing on 32 years of pharmacy practice. I take on a small number of projects at a time so each one gets full attention.
Best fit: teams that value getting clinical logic right over shipping fast.
Email Paul