Nirajan K.

Backend engineer first, AI engineer second. I ship production systems that work. RAG platforms, FastAPI, Redis. I write clean code and architecture that lasts. Top Rated, 100% Job Success. I know what I'm doing, and I make it easy for you to get what you need. Let's get to work. What I build - RAG over your documents and databases: pgvector, Qdrant, Pinecone, or Weaviate, hybrid search, re-ranking, and RAGAS evals so answer quality is measured, not guessed - AI agents and multi-agent workflows with tool calling, memory, and human approval steps (LangGraph, LangChain, Claude, OpenAI, Gemini) - MCP servers that give Claude and other agents safe, scoped access to your internal tools and data - LLM features inside existing products: streaming chat, structured outputs, function calling, prompt caching, token and cost tracking with Langfuse - API integrations and AI workflow automation in code: webhooks, Stripe, HubSpot, Slack, Airtable, Supabase, or n8n, Make and Zapier when your team already runs on them Why my AI systems reach production Before AI work, I spent years building and running Python backends and full-stack products. That is the layer most AI prototypes skip. - 12,000 RPM at 99.99% uptime after a zero-downtime Django monolith-to-microservices migration - 72% lower API latency from an async FastAPI and Redis redesign - 2M+ events/day at 37ms average latency on a containerized pipeline with PostgreSQL sharding - Release cycles cut from 2 days to 2 hours with GitHub Actions CI/CD - 35% lower AWS infrastructure cost through rightsizing - On a recent Anthropic API build, I audited the backend unprompted and flagged unauthenticated read routes as the first fix. The client verified the call months later. Full-stack, when the AI needs a product around it - Backend: Python, FastAPI (async-first), Django, DRF, Celery, WebSockets, PostgreSQL, Redis - Frontend: React, Next.js App Router, TypeScript, Tailwind - Cloud and DevOps: AWS (ECS, RDS, S3, Lambda), Docker, Kubernetes, GitHub Actions, Prometheus and Grafana How I work - A short written plan before code: architecture, risks, and expected cost per request - Milestones that end in a working demo, with tests and CI from the first one - Plain written updates and a handover your team can run without me - Based in Nepal (UTC+5:45), with daily overlap for Europe and US East mornings Not a fit if - You need a chatbot wrapper shipped tomorrow with no tests or monitoring - The budget covers a prompt, not a system - You need model training or computer vision research. I build applied LLM systems on top of foundation models Send me the workflow or product you want AI inside, plus what it costs you today. Already have a RAG or agent system that is live but wrong? I will audit it first for a fixed price, then give you the main risks and a fixed-price first milestone.