Remote: Yes
Willing to relocate: Yes
Technologies: Node.js, Express.js, Next.js, TypeScript, JavaScript, Python, PostgreSQL, Redis, BullMQ, Prisma, Socket.io, React, Docker, AWS, RAG Pipelines, agentic AI, LangChain, LangGraph, Pinecone, LLM Integration
Résumé/CV:https://drive.google.com/file/d/1TQtVpiRiStBEMlSMkjr6o7hZaDp...
Portfolio: https://praveen-portfolio-dev.vercel.app/
Email: praveenprabhakarkumar@gmail.com
I keep turning single-tenant hacks into systems that scale to N — without a rewrite.
In my last role, I took a Fortune 50 telecom client's warehouse platform from a hardcoded single-store system to a config-driven multi-store architecture, taking daily EDI throughput from ~200 to 500-600+ transactions/day with support for unlimited future store onboarding. Same instinct shows up elsewhere I've worked: an async forecasting engine that now powers $30-40M revenue and $10-20M labour projections across 10-12 warehouses, and a geospatial load-matching engine (PC Miler + DAT APIs) that cut empty driver miles from 35%+ to under 25%.
Recently I've been applying the same discipline to AI products. Memorix (Live: https://memorix-sepia-kappa.vercel.app/) is a multi-tenant RAG platform , Pinecone namespace partitioning and Supabase RLS for isolation, two-stage retrieval/reranking, OAuth into GitHub/Slack/Notion/Confluence.
NextFlow is a DAG-based AI workflow engine (React Flow + Trigger.dev) where independent nodes run concurrently via topological sort, with branch-isolated fault tolerance so one failure doesn't take down the rest.
Currently going deep on AI agents and MCP — interested in roles where I can take that from prototype to production.
Looking for backend, full-stack, or applied AI/GenAI engineering roles.