I build AI products, agent systems, RAG pipelines, and backend infrastructure with a strong focus on how things work under the hood.
Current focus → GenAI + Agentic AI + Backend Engineering
Building → AI agents, RAG systems, APIs, async pipelines, developer tooling
Main stack → TypeScript, Bun/Node.js, React, PostgreSQL, Redis, Docker, AWS
I like building systems that are practical, inspectable, and production-oriented — not just demos. Most of my work sits at the intersection of AI application engineering and backend infrastructure.
|
Autonomous developer agent harness A TypeScript/Bun agent that can inspect repositories, edit code, execute sandboxed commands, work with Git, research on the web, and iterate across multi-step tasks. Highlights
|
Transparent TypeScript agent SDK A small provider-agnostic SDK built around an explicit Highlights
|
|
Multi-source RAG platform Ingests PDFs, websites, YouTube transcripts, and raw text, then answers questions using grounded retrieval over pgvector-backed embeddings. Highlights
|
AI resume analysis platform Processes resumes asynchronously and returns AI-generated analysis while keeping the request path fast through queues, caching, and persistent storage. Highlights
|
| AI / GenAI | RAG · Vector Search · AI Agents · Tool Calling · Structured Outputs · Prompt Engineering · Guardrails |
| Backend | REST APIs · API Design · Postman · Auth · Background Jobs · Rate Limiting · Queues · Caching · File Pipelines |
| Data | PostgreSQL · pgvector · Redis · MongoDB · Prisma · SQLite |
| Infrastructure | Docker · AWS · Vercel · Render · Linux · GitHub · Bun · Node.js |
- Agent architecture and orchestration
- RAG quality, retrieval evaluation, and grounding
- Reliable async processing and job systems
- API gateways, quotas, rate limits, retries, and fault tolerance
- System design through practical implementation
- AI infrastructure and developer-facing AI tools



