Agentic AI SystemsThat Hold Up in Production
Agentic Architectures
Production-grade RAG, orchestration, and workflow design
Evals & Reliability
Validators, monitoring, regression thinking, and CAT
High-Stakes Domains
Healthcare, legal, commerce, identity, and payments
Featured Work
Public writing, production case studies, and systems work centered on agentic architecture, evaluation, and reliability engineering.
Reliability Testing for LLM-Based Systems
The CAT Framework White Paper
Comprehensive framework for conducting reliability tests on LLM systems using validators, verifiers, and reliability tensors. Includes binomial experiments, generative conditional validators, and production monitoring strategies.
Agentic Architecture
Structuring Agent-Tool-User Interactions
A structured approach to modeling interactions between agents and tools through chat histories, cycles, and graph-theoretic representations. Foundational concepts for understanding agentic AI systems.
Formalized Structures: The Algebra of Agentic Architectures
Mathematical Formalization of Tool-Using AI Systems
Mathematical formalization of agentic AI through tool call matrices, selection masks, and tensor operations. Defines the algebra underlying agent-tool interactions with formal propositions and proofs.
eBay
2025
Enterprise-scale Agentic AI system optimizing legacy seller workspaces.
AI Capabilities
Key Highlights
- Architected agentic AI workflows for enterprise-scale deployment
- Integrated AI capabilities into legacy seller workspace systems
- Deployed production AI at massive scale
Mayo Clinic
2024
Multi-agent RAG system accelerating medical research and discovery for one of the world's leading healthcare institutions.
AI Capabilities
Key Highlights
- Designed and implemented multi-agent architecture for medical research
- Built retrieval-augmented generation (RAG) system for knowledge synthesis
- Enabled researchers to accelerate discovery through AI-powered insights
Trust & Will
2024
Attorney-in-the-loop automation system digitizing estate planning practices with human oversight and AI efficiency.
AI Capabilities
Key Highlights
- Designed human-in-the-loop AI architecture for legal workflows
- Automated estate planning document generation and processing
- Maintained attorney oversight while increasing efficiency
Arrive Health
2024
Complex AI flows condensing critical clinical information to support patient care and healthcare decision-making.
AI Capabilities
Key Highlights
- Built AI systems for clinical information processing
- Designed complex information extraction pipelines
- Enabled AI-driven summarization for critical patient support
Graph Theoretic Multi-Agent Dynamics
My First Python Script
Mathematical simulation of autonomous agents reaching consensus through graph-based communication. Graduate thesis work in Applied Mathematics.
Hillcrest Ski & Sports
Built and continuously maintain e-commerce platform driving over $1M in annual sales. Ongoing client with active support and feature development.
Key Highlights
- Built company's first dynamic e-commerce platform (2020-2021)
- Transformed static site into real-time multi-channel fulfillment system
- Integrated 23,000+ SKUs with automated data pipelines