WHAT I’M BUILDING

Multiple Projects, One Mission.

Govafy is where I put my government contracting experience into software. Fusion Coding grew from the process of building it. These case studies show my architecture decisions, development workflow, and what I’m learning along the way.

Platform architecture

Govafy

An experimental AI platform I architected to connect government contracting data and institutional knowledge with opportunity discovery, document processing, and proposal workflows.

Architected the platform with TypeScript, React, Encore.ts, and PostgreSQL/pgvector
Built data ingestion for more than 30 million award records, alongside document processing with OCR
Connected keyword and vector retrieval for RAG with source traceability and human verification

What I built

Govafy explores opportunity discovery, proposal drafting, contracting officer research using keywords and NAICS codes, and a pipeline dashboard for procurement activity.

My contribution

I defined the contracting workflows and architected the platform. My work spans requirements, workflow design, implementation, testing, debugging, and cloud deployment.

What I learned

Building Govafy taught me to distinguish between storing information, making it searchable, and retrieving context that helps answer a question. Working with award records and documents deepened my understanding of data quality, access boundaries, source traceability, and end-to-end workflows.

Project Details ↗

AI-assisted development

Fusion Coding

An AI coding project I created while building Govafy, exploring how competing agents, structured evaluation, and human oversight can improve software quality and pull request reviews.

Created tournaments of competing AI agents with independent evaluation and integration
Explores repeatable planning, implementation, testing, review, and verification workflows
Compares implementations against a shared specification and clear acceptance criteria

Why I built it

Generating code was only part of the challenge of building Govafy. I needed a repeatable way to plan work, compare implementations, review changes, and verify that the result met the requirements.

My contribution

I created an AI meta-harness that coordinates competing agents from different frontier labs. I define specifications and acceptance criteria, work through implementation and debugging, and evaluate the results. A tournament can produce a single winner, no clear winner, or an integration of selected components.

What I learned

The project has deepened my understanding of TypeScript, agent orchestration, testing, Git branches and worktrees, and software delivery. It has taught me to separate a convincing explanation from a working implementation and make success specific enough to evaluate.

Project Details ↗

THE BRIDGE

I own the architecture and decisions.

AI tools are part of how I build.

I use Linear to organize requirements, dependencies, and acceptance criteria, then work with Claude Code, Codex, and Cursor to implement and refine the software. Fusion Coding tournaments compare approaches before integration. My work spans testing, debugging, code review, version control, deployment, and checking the result. Each project documents the technical choices and lessons behind that process.

Connecting customer understanding with technical execution.

These projects strengthen my ability to translate customer workflows into technical requirements and working demonstrations. That experience is relevant to government sales, customer-facing AI work, and forward-deployed engineering.

I measure progress through working functionality, informed technical decisions, and what I can demonstrate, explain, verify, and improve.