01 / Developer toolingPUBLIC REPOSITORY · ALPHA
ecc-init
Set up rollback-friendly Claude Code configuration in an existing project
- Python
- Claude Code
- CLI
- Rollback
- 149
- tests passed
- 3
- OS CI matrix
FULL-STACK DEVELOPMENT · AI PRODUCTS
I build AI products that work beyond the demo.
I work mainly with Java, Vue, and Python, bringing RAG, MCP, and language models into real product workflows. I care about the data, tests, deployment, and recovery path—not just the model call.
SELECTED WORK
Projects explained through decisions, trade-offs, and evidence.
02 / Research toolingPUBLIC REPOSITORY · ALPHA
GroundedSeek
Keep every research conclusion traceable to its evidence
- FastAPI
- MCP
- Playwright
- SQLite
03 / Full-stack applicationPUBLIC REPOSITORY · ALPHA
PhoneMall
A self-hosted commerce platform with a RAG shopping assistant
- Spring Cloud
- Vue 3
- RAG
- MySQL
- 4
- backend services
- 39
- backend tests
CAPABILITIES
From product requirements to a running deployment.
Full-stack application development
I take products from domain models, APIs, and user interfaces through database migrations, containers, and self-hosted deployment.
- Java
- Spring Cloud
- Vue 3
- Python
- Docker
RAG and intelligent search
I connect models to real data and build retrieval, citations, streaming responses, and evaluation into the workflow.
- RAG
- MCP
- SSE
- Vector Search
Quality and operations
Tests, CI, logs, backups, and rollback keep a project maintainable after the first successful demo.
- CI/CD
- Testing
- Audit
- Rollback
HOW I WORK
Turning an idea into a product without losing control of it.
- 01
Define the real problem
Clarify the user need and identify which decisions must remain human.
- 02
Complete the workflow
Connect the interface, APIs, data, model, and deployment environment.
- 03
Check the result
Use automated tests, static checks, and source records to catch mistakes.
- 04
Plan for recovery
Keep important changes traceable and make failures recoverable through backups and rollback.
LATEST WRITING
Notes from problems I have actually encountered while building.
Building an Evidence-backed Web Research MCP
Moving from search results to reviewable conclusions requires source snapshots, claim links, conflict handling, and versioned artifacts—not just a browser tool.
- MCP
- Web Research
- Verifiable AI
Why Long-form AI Writing Needs a Canon Boundary
Models can generate plot ideas and prose quickly, but long-form work stays coherent only when suggestions, drafts, and accepted facts remain separate.
- AI Writing
- Human-in-the-loop
- Architecture
ABOUT EXECUTE42
Putting AI into products—not just into chat boxes.
I use this site to document projects, architecture decisions, and engineering lessons. Start with the case studies, or visit GitHub for the public code.