AI agent systems, automation, and retrieval for non-healthcare domains
The same engineering and advisory work from Services and Consulting — applied to industry, operations, and enterprise automation problems.
What this covers
The JTMDAI lab does three things: AI agent and automation systems, secure self-hosted tooling, and physician + AI advisory. All three translate directly to non-healthcare industry work — the same engineering rigor, the same bias toward auditable and sovereign systems, the same insistence on citing your sources and not fabricating claims.
The Medical RAG retrieval engine is one instance of a broader RAG architecture: hybrid retrieval that runs dense semantic search over text embeddings alongside classic keyword (lexical) search, then merges the two ranked lists with reciprocal rank fusion. The same corpus-indexing and retrieval pipeline can be pointed at industry-specific documentation, regulatory filings, internal knowledge bases, or any structured body of text where traceable, cited answers matter more than fluent confabulation — and retrieval quality is measured with standard information-retrieval metrics (recall@k, nDCG), not asserted.
Focus areas
Multi-agent workflows that coordinate, reason, and act with clear boundaries on what each agent can touch. Reasoning runs on large language models through a ReAct loop (reason → act → observe), local-first with a cloud fallback. The same architecture that powers Guardian One — finance, scheduling, inbox, home, security coordination — adapted to your operational domain. See Services for detail.
Local-first, encrypted, audit-logged systems that the client owns end to end. Credentials stay encrypted, reasoning runs on infrastructure you control when it can, and every meaningful action lands in an append-only record. No black boxes, no quiet data exploitation. See Services.
The same hybrid retrieval pipeline built for public medical literature — dense embeddings + lexical search, fused and re-ranked — can index regulatory filings, technical standards, internal documentation, or policy documents, returning cited, traceable answers rather than fabricated summaries. Based on the same architecture as Medical RAG.
A clinician’s rigor applied to AI and automation: thinking through risk, evidence, and edge cases the way medicine demands. The advisory ranges across the working algorithm toolkit — retrieval and ranking, embeddings and similarity, classification and scoring, clustering, recommendation, and graph analysis — matching the right (often boring) technique to the problem instead of reaching for a model that has to be trained, hosted, and trusted. Useful well outside healthcare, wherever the cost of an AI system being wrong is high. This is engineering and advisory work, not medical advice. See Consulting.
Engagement model
Engagements start with a deposit applied to the work, and are scoped via an engagement letter before any material work begins. The Consulting page describes the available package tiers — from a 14-day retainer to a fixed-scope audit sprint to fractional advisory. Industry engagements follow the same structure.
The lab takes on a small number of engagements at a time. If you have a scoped problem in AI, automation, or retrieval for an industry domain, reach out with what you are building and what you are trying to solve.
Tell me what you are building and what you are trying to solve. The lab takes on a small number of engagements — AI agents, automation, retrieval, and secure tooling for clients who want traceable, auditable systems, not black boxes.