AI + Complex Systems Architect for Regulated, Chaotic Environments

Joshua Farrow builds AI platforms that survive audits, adverse events, and executive scrutiny—turning ambitious LLM ideas into reliable, observable, and secure systems.

Assured AI Architecture
Governed Data & Telemetry
Risk-Aware Product Strategy

Safety-Critical Design

Blueprints agentic systems with containment, observability, and fail-safes for regulated operations.

Reliability & Monitoring

Ships telemetry, red-team signals, and executive-ready metrics that keep AI decisions explainable.

Executive & Board Alignment

Translates regulatory, reputational, and business constraints into actionable AI roadmaps.

Schedule Architecture Call View Portfolio

Overview & Approach

AI/LLM architect for chaotic, regulated environments who turns seeminlgy unstructured, multi-actor workflows into governed, measurable systems—always aligning incentives, compliance guardrails, and operational realities.

Philosophy Design for reality: every interaction is logged, schema-bound, testable, and considerate of policy, billing, and operational constraints.
Architect & translator Sits between domain experts and engineers—mapping incentives, surfacing lost information, and prototyping enough of the stack (Python, SQL, APIs, LLM orchestration) to give every stakeholder a concrete blueprint.
How to Engage Bring regulated, multi-stakeholder workflows or ambitious AI bets; I’ll map the system, design safe RAG/function calling plans, wire evaluation and KPI loops, and deliver prototypes that withstand audits and executive scrutiny.

What I Do

Partnering across discovery, design, and delivery to align LLM architecture with regulated operations.

LLM Architecture in Regulated Contexts Map multi-actor workflows into schema-constrained agents with grounded RAG, tool/function calling, refusal handling, and HIPAA-aware telemetry.
Workflow Mapping & KPIs Decompose clinical and operational journeys, define metrics/KPIs, and wire data contracts so adoption, quality, and cost deltas are visible.
Prototypes & Cross-Functional Translation Build Python/React proofs with evaluation harnesses while translating between frontline teams, engineers, compliance, and leadership.

Selected Outcomes

Evidence from regulated LLM systems, workflow mapping, and KPI-driven delivery.

2024 documentation copilot Architected RAG + function-calling assistant with schema-bound plans/notes and eval harnesses; documented ~60% time reduction in ABA pilot workflows while satisfying HIPAA logging.
Cross-functional workflow maps Mapped diverse behavioral record and revenue cycle flows into shared service blueprints that aligned enterprise engineering, quality, and stakeholder requirements.
Evaluation-led prototypes Delivered Python/React proofs with structured prompts, safety tests, and KPI dashboards to de-risk launches and shorten adoption cycles.

Expertise

What teams rely on when aligning AI architecture, compliance, and measurable operations.

LLM & AI Architecture Schema-validated tool use, grounded RAG, evaluation harnesses, refusal/uncertainty patterns, and telemetry designed for audits.
Python & Data Systems Pandas, NumPy, Pydantic, FastAPI prototypes, ETL across PDFs/CSVs/APIs, Azure Data Explorer, PostgreSQL, KPI instrumentation.
Workflow & UX Prototyping Web app prototypes that keep operations, compliance, and technical teams aligned and informed.
Compliance & Risk Alignment HIPAA/ePHI minimization, RBAC, audit-ready documentation, de-identification, and telemetry patterns adopted in healthcare and regulated industries.
Regulated Operations Experience across healthcare, behavioral health, education, and payer-driven ecosystems with a focus on making complex systems measurable and reliable.

GitHub Projects

Open-source experiments that stress-test AI, complex data, and real-world constraints end-to-end.

n-dimensional-llm Research exploration of multi-field information bottlenecks and rate–distortion theory for structured data compression and high-dimensional world models.
semantic-tensor-analysis A temporal semantic evolution analysis framework for tracking how meaning changes across time, tokens, and context.
qca_test_qatnu A custom many-body simulation engine for the QATNU/SRQID quantum theory model, running parallel λ scans on arbitrary graph topologies and exporting figures/logs for 2025 preprints.
glass-bead-game An LLM-based recreation of Hermann Hesse’s Glass Bead Game, exploring AI-judged, multi-domain idea weaving and adjudication.
governable-agi-charter A copy-ready governance starter kit for deploying advanced models under verifiable constraints, focused on alignment, auditability, and control surfaces.
gamma_studies Options and futures analytics utilities (with related repos futures_trade_analysis and gamma_db_analysis) for measuring market structure, risk, and gamma exposure from real trading data.
algorithm-mirror A Chrome extension that uses local vision and text models to analyze your personal content feed, surfacing how algorithms are shaping what you see.

Explore the full GitHub profile

Research, Publications & Service

Contributions spanning AI systems, reinforcement learning, quantum information theory, and applied sciences.

Co-author “Reinforcement Learning in Behavior Science and AI” (forthcoming book chapter).
Prior lab work ARVO abstract in IOVS on proton-coupled folate transporter mechanisms relevant to retinal health.
Peer reviewer Journal of the Experimental Analysis of Behavior (JEAB) and The Bulletin (AI/LLM research).

Pre-prints

Speculative research on quantum information dynamics and computational universes.

Emergent Physical Laws from Self Referential Quantum Information Dynamics (SRQID) SSRN pre-print exploring how self-referential quantum information dynamics may give rise to physical laws.
Quantum Automaton Tensor Network Universe (QATNU) SSRN pre-print proposing a tensor network universe model grounded in quantum automata perspectives.

Background & Credentials

Foundation across applied sciences, behavior science, and regulated operational environments.

Professional experience 15 years operating within highly regulated care and education systems, later transitioning into AI systems architecture.
Education M.S. Counseling (ABA concentration), Nova Southeastern University; B.S. Biochemistry, Virginia Tech (Biotechnology minor).

Let's Collaborate

Share your goals for regulated AI, analytics, or workflow redesign — I respond within two business days.

Email info@joshuafarrow.com
Calendar Book an intro session to explore alignment.
Social LinkedIn · GitHub