Most jobs reach me the same way. A founder has an app a coding agent built, it demos fine, and it breaks the week real users arrive. I go in and add what was never there: tenant isolation, working auth, migrations that don't lose data, tests, and enough logging to find out what broke.
The other half of my work is building the agents themselves — retrieval that cites its sources, tools with permissions the model can't talk its way past, and evals so I know when a prompt change made things worse.
Backend is NestJS and Node on Postgres. Frontend is React and Next.js when a project needs one.
Claude Certified Developer – Foundations (Anthropic)
Earned November 4, 2026 · Expires November 4, 2027.
Validated knowledge of Claude API mechanics, agent architecture and patterns, context engineering, tool implementation, MCP development, AI application security, and token/cost management.
Note
Open to contract work. [email protected] · LinkedIn
prod-readiness — Claude Code plugin that audits a repo before launch. Seven agents review security, database, DevOps, QA and AI-security over one shared evidence pass, then return SHIP, FIX THEN SHIP, or HOLD.
boring-react — Agent Skills that stop coding agents writing clever React. No needless useEffect, no memo on everything, no fresh dependency to save four lines. 12 adversarial evals check it holds when a prompt pushes the other way.
langraph_agent — LangGraph agent on NestJS. Classifies each query, routes it to RAG, chart or hybrid, runs retrieval per tenant against Weaviate, streams the answer back with page references.
ecommerce-backend — Node, Express, GraphQL and Postgres.
Also: LangGraph, LangChain, pgvector, Weaviate, MCP, Claude Code, Codex.
tamork.tech · LinkedIn · X · Stack Overflow · [email protected]