Matt Coburn
AI Evaluation, Agent Governance and Enterprise AI Leadership
I'm an engineering leader with 11 years across machine learning, production AI and evaluation who still builds. I put AI agents into regulated production and show they work, with evaluation-based release gates, human-in-the-loop controls and auditable decision trails.
I led technical direction for an enterprise AI platform deployed at 2 of the top 3 global automakers, founded an enterprise AI business that generated ~$2M in revenue, and led teams of up to 15 data scientists as VP of Data Science. I'm currently building an agent-monitoring system that keeps people aware and in control of what AI agents are doing in their organizations.
Independent AI Safety Research and Engineering
Jul 2026 – Present
- Building an agentic system for monitoring AI safety that keeps people aware and in control of what AI agents, and the people using them, are doing across an organization
- Researching AI privacy and AI cyber security, including red teaming and jailbreaking
Digital Ascension Group (DAG) — Head of AI Engineering
Mar 2026 – Jun 2026
- Agent governance: In under four months, established governance for AI agents and designed and implemented the compliance process for organizational AI usage
- Model strategy: Defined the AI model strategy
- Data platform: Architected and built the company's end-to-end data platform, including data governance and operational data infrastructure for sales and marketing
Aristotle — Founding Engineer & Tech Lead
Los Angeles, CA | Jun 2024 – May 2026
Led technical direction for an enterprise AI platform deployed at 2 of the top 3 global automakers, automating complex business workflows.
- Agentic AI architecture: Built tool-using agents as explicit state machines with boundary validation, retry and escalation logic, and human-in-the-loop controls
- Validation before action: Enforced validation before outputs reached users or changed system state; shipped document extraction with schema validation, click-to-source provenance and auditable decision trails
- Team leadership: Hired and led 4 engineers; established CI/CD, code review, observability and evaluation-based release gates
- Hybrid retrieval & evaluation: Tuned hybrid BM25 and dense retrieval using Precision@K and Recall@K, with regression testing required for release
WorkFusion — VP of Data Science
New York, NY | May 2023 – Oct 2023
Led ~15 data scientists and owned the applied AI roadmap for KYC/AML systems, document extraction and screening in regulated financial services.
- Compliance AI at scale: Improved detection outcomes by ~35% by shipping transaction monitoring, risk scoring and case prioritization for major financial institutions
- Responsible AI operations: Designed security-first MLOps, evaluation, bias mitigation, model monitoring and audit practices for production systems
- AI screening & document intelligence: Built LLM-based extraction and auto-labeling for high-volume screening workflows with governance and explainability requirements
Tangible Intelligence — Founder & CEO
Dallas, TX | Jan 2020 – Apr 2023
Founded an enterprise document intelligence company, generated ~$2M in revenue and led an 8-person engineering and data science team.
- Agents at production volume: Built agents that processed 300,000 paystubs per month for MasterCard mortgage applications, supporting downstream delivery of extracted data to Fannie Mae and Freddie Mac
- Enterprise operations: Delivered SOC 2–certified deployments and Fortune 100 integrations; created a no-code, human-in-the-loop extraction platform with ontology-driven schemas and source provenance
- Evaluation before delivery: Designed randomized human-versus-agent comparisons to evaluate information extraction before downstream delivery; built the UX, models and evaluation process about two years before ChatGPT
- SafeScan: Built a sensitive-data detection and anomaly-screening product combining ML, rules and search, generating ~$900K in standalone revenue
M Science (Jefferies subsidiary) — Principal Data Scientist, Founding Lead
New York, NY | May 2018 – Dec 2019
Built an internal AI venture at the Jefferies subsidiary, conceiving and shipping an analytics platform for institutional clients including BlackRock, Two Sigma, and Citadel.
- Client-facing AI product: Enabled clients to apply shared analytics to private datasets with isolation and fine-grained access control; designed reproducible entity resolution
- Enterprise analytics platform: Created an Ontology-as-a-Service layer for companies, securities, events and relationships across public and proprietary data
Expedia Group (Hotels.com) — Data Scientist
Dallas, TX | Oct 2016 – May 2018
- Revenue optimization at scale: Influenced $200M+ in annual advertising spend with click-value models across millions of keywords that enabled automated bid optimization
- ROI & spend allocation: Delivered ~240% ROI improvement on targeted segments by identifying underperforming publisher inventory and reallocating spend; built production monitoring and safeguards
Evaluation and Validation Methods
Execution-trace and trajectory evaluations, scoring rubrics and human/agent comparisons. LLM-as-judge methods and AI-assisted, partially synthetic evaluation datasets.
Additional Experience
Data privacy and security: Built a PII identification tool for Spirion.
Technology Stack
AI & platform: Claude, LangChain, LlamaIndex, PydanticAI, pgvector, Pinecone, OpenRouter, Microsoft Azure, Google Cloud Platform (GCP), Databricks, React
Core: Python, SQL, TypeScript, C, Zig
Backend & Infrastructure: FastAPI, Pydantic, PostgreSQL, Redis, Docker, Kubernetes, AWS
Data & Analytics: Pandas, NumPy, Spark, ETL/streaming pipelines, vector databases
AI/ML: PyTorch, HuggingFace, OpenAI/Anthropic APIs, agentic frameworks
Specialties: AI evaluation and validation, human-in-the-loop agent control, agent governance and compliance, red teaming and jailbreaking, regulated AI (KYC/AML, SOC 2), enterprise SaaS, document intelligence, compliance AI, structured extraction, screening systems, AI governance, responsible AI, production machine learning, evaluation frameworks
Education
University of Texas at Dallas
- B.S. - Electrical Engineering, 2013
- Graduate Coursework - Computer Science, 2016
Availability
Available immediately for VP-level AI leadership roles. Email [email protected].