Joshua Farrow
AI Engineer + Systems Architect for Healthcare Operations
I build AI systems for clinical workflows—documentation, intake, and compliance monitoring. After 15 years inside behavioral health operations, I know where the real bottlenecks and audit failures come from. I also build the infrastructure to fix them: local-first, testable, evidence-producing systems that clinics can actually deploy.
Built and shipped: agentic clinical intake system piloted by 50 clinicians across 15 ABA agencies—60%+ reduction in documentation time.
Currently: exploring pilot partnerships for clinical NLP infrastructure (on-prem / no data egress).
Previously: built an agentic clinical intake system at Neuromnia—RAG pipeline with Semantic Kernel that processed audio recordings and uploaded documents into draft intake summaries and structured clinical report sections.
Or: run a 500-session STM pilot (on-prem, no data egress)
What I Build
- Clinical AI Systems RAG pipelines, document processing, structured report generation—designed for HIPAA environments with audit trails and strict data boundaries.
- Workflow Automation Turn unstructured clinical text into measurable signals: note quality checks, intake extraction, treatment-plan consistency, and documentation QA.
- Compliance-Ready Infrastructure Local inference, air-gapped deployment options, evidence-grade outputs (reports + JSON + provenance) that survive audits and peer review.
Selected Work
- STM Engine — Semantic Analysis for Clinical Documentation A deployable microservice that measures semantic drift in clinical notes over time. Flags documentation inconsistencies, highlights risk signals, and generates audit-ready evidence reports. Runs on-prem in Docker—no data leaves your infrastructure. Outputs: evidence report (PDF/HTML) + metrics JSON + provenance manifest.
- Neuromnia — Agentic Clinical Intake System (2024) Built a system that processed audio recordings and uploaded documents into draft intake summaries and structured clinical report sections. RAG pipeline with Semantic Kernel orchestration. Piloted by 100+ clinicians across 10 ABA agencies with 60%+ reduction in documentation time. Founding engineer and CPO.
- Mnemolog — Conversation Archive Platform Full-stack SaaS built in 48 hours: Supabase auth, consent-based upload flow, PII flagging, and a public archive experience. Proof of rapid 0→1 execution.
GitHub
- semantic-tensor-analysis Temporal semantic drift measurement for documentation QA.
- mnemolog Production archive + consent/review + sensitive content handling.
- n-dimensional-llm Structured compression techniques for high-dimensional clinical embeddings.
Background
- Operations experience 15 years in behavioral health operations—clinical work, supervision, program management. I've written the notes, run the sessions, dealt with the audits.
- Education M.S. Counseling (ABA concentration), Nova Southeastern University; B.S. Biochemistry, Virginia Tech.
- Credentials Board Certified Behavior Analyst since 2014.
Publications & Review Work
- Co-author "Reinforcement Learning in Behavior Science and AI" (forthcoming book chapter).
- Peer reviewer Journal of the Experimental Analysis of Behavior.
Contact
If you're building AI for healthcare operations—clinical documentation, intake, QA/compliance, or behavioral health workflows—and you need someone who can bridge domain reality and engineering execution, let's talk.
Pilot option: run a 500-session STM pilot (on-prem, no data egress) — deliver evidence reports + metrics JSON + readout.
- Email [email protected]
- Calendar Book a call
- Social LinkedIn · GitHub