Case Studies

Selected Projects

Real systems built for real businesses. Each project is a production deployment — not a demo, not a prototype.

Regulatory AI · Compliance · Multi-Service · Agentic

WCP Compliance Agent

Federal payroll validation with a three-layer decision pipeline

Built a compliance system for Davis-Bacon Act payroll validation that answers one question contractors actually fear: "Can we defend this in a DOL audit?" Three layers of proof — arithmetic, AI review (must cite statute), trust scoring — with 413 tests and every decision traceable to the regulation.

  • 413 tests across V2 and V3 — deterministic validation, not LLM guesswork
  • Every decision cites 40 U.S.C. § 3142 with immutable audit trails
  • V3 architecture: React 19 + TypeScript agent + Python/FastAPI backend
  • Three-service split designed to survive regulatory review
Read more
Algorithms · Frontend

Valid Palindrome

Algorithm visualization that makes complexity visible

A LeetCode 125 showcase with three algorithmic approaches — two-pointer, reverse-and-compare, recursive — visualized through a retro terminal interface. Tracks live metrics, step-by-step execution, and compares space/time tradeoffs side by side.

  • Three approaches with full complexity analysis and tradeoff comparison
  • Retro terminal aesthetic: CRT scanlines, phosphor glow, animated execution
  • Live metrics: time, space, allocations, comparisons, early exits
  • Custom input and adjustable speed controls
Read more
Automation · API · Evaluation

Model Router Skill

Production LLM routing that never leaves you stranded

A production skill for Hermes Agent that routes every task to the right LLM provider — balancing capability, cost, and availability across 5+ providers with automatic fallback chains. Keeps agents running when APIs fail, and saves money by matching task complexity to model capability.

  • Dynamic provider selection with cost optimization and capability-based routing
  • Automatic fallback chains — when a provider fails, it recovers without human intervention
  • Real-time health monitoring and availability tracking
  • Version 4.4 — iterated based on real production usage
Read more
Automation · LLM · Agents

AI Content Systems

Multi-agent pipelines that research, draft, and repurpose at scale

End-to-end automation pipelines where a research agent gathers context, a draft generator produces long-form copy, and format adapters transform it into threads, newsletters, and summaries. Built around your knowledge base — not generic templates.

  • Multi-agent workflow: Research → Draft → Repurpose → Publish
  • Format adaptation: blog → Twitter thread → LinkedIn → newsletter
  • Human-in-the-loop approval gates for quality control
  • Your knowledge base, not generic templates
Read more