AI Developer @ erbrains | Building production LLM applications & AI systems
- Applied AI at erbrains: LLM application development, RAG systems, AI agent workflows
- Stack: Python, LangChain, LlamaIndex, FastAPI, Vector Databases (Pinecone/Weaviate/pgvector)
- Transitioning from: 3+ years production Frontend Engineering → Applied AI
- Frontend Lead capabilities: React, Next.js, TypeScript, React Query, Redux, RxJS
- Architected & shipped at scale:
- Multi-tenant healthcare platform (5+ tenants, isolated deployments, RBAC)
- Real-time EV fleet tracking (1,000+ vehicles, OpenLayers, marker clustering)
- Server-side PDF generation with custom pagination & encryption
- Device-specific responsive systems (component-level, not just CSS)
- React Query expertise: True optimistic updates w/ rollback, typed wrapper layer, centralized query-key factory, cursor/page infinite scrolling
| Project | Stack | Highlights |
|---|---|---|
| Social Connect | Next.js 15, TypeScript, Prisma, PostgreSQL, Socket.io, Docker | Full-stack social platform. True optimistic updates w/ rollback (React Query), cursor-based infinite feeds, real-time messaging via separate Socket.io microservice |
| Highlight Saver Extension | TypeScript, Chrome Ext MV3, esbuild, Gemini API | Shadow-DOM UI injection, local-first storage, on-demand AI summaries |
Note: AI projects from my work at erbrains are proprietary. This profile demonstrates the engineering rigor I bring to AI development — clean architecture, typed systems, caching strategies, and production-grade delivery.
AI/ML (Current): Python, LangChain, LlamaIndex, OpenAI/Anthropic APIs, Hugging Face, RAG, Agent Patterns, Vector DBs
Backend: FastAPI, Node.js, Express, PostgreSQL, MongoDB, Redis, Prisma
Frontend (Deep): React, Next.js, TypeScript, Tailwind, React Query, Redux, RxJS, Framer Motion
DevOps/Tools: Docker, GitHub Actions, Git, Linux, VS Code
- Production mindset: 3+ years shipping user-facing SaaS at scale
- Type-safe by default: TypeScript → Python type hints, Pydantic, strict configs
- Caching & state expertise: React Query patterns transfer directly to LLM response caching, conversation memory, tool call optimization
- System design: Multi-tenant architecture, RBAC, isolated deployments — relevant for multi-user AI apps
- Learning velocity: Self-taught, bootcamp grad, consistently shipped complex features
- 💼 LinkedIn: imdebabrata
- 🌐 Portfolio: imdebabrata.github.io
- ✉️ Email: debabratakgt@hotmail.com
- 🐙 GitHub: @ImDebabrata
Currently building AI systems at erbrains. Open to conversations about applied AI, LLM app architecture, and full-stack AI product engineering.



