Muhammad Nur S.

I build AI automation that actually ships to production — RAG assistants, AI agents, and AI-powered Telegram bots. Ranked 19th of 1,000+ teams at the AI 500 Hackathon Central Asia, and I build the real systems behind the AI, not demos. Most "AI automation" freelancers stop at no-code tools (Zapier, Make, GPT wrappers) that break the moment real traffic hits. I'm a backend engineer first, so I build the actual system underneath: proper database schema, async APIs, vector search, queues, auth, and Docker deployment — the foundation your AI product needs to survive real users. What I build for you: - AI Agents & Automation — OpenAI-powered agents and tool-calling workflows that connect to your real data and APIs. - RAG Assistants — production Retrieval-Augmented Generation with pgvector / HNSW and OpenAI embeddings. I've built a source-grounded assistant over 150,000+ documents with hallucination-resistant prompting. - AI Telegram Bots — smart assistants on Aiogram 3 with AI replies, payments, webhooks, admin panels, and background jobs (Celery + Redis). - Production Backends — FastAPI / Django REST APIs, PostgreSQL, JWT auth, fully Dockerized and deployed with CI/CD, nginx + SSL. Proof, not promises: - AI 500 Hackathon Central Asia — 19th of 1,000+ teams (applied LLM/RAG under time pressure). - Vakilim.ai — a multi-tenant RAG AI assistant with Telegram + admin panel, live in production. - Backend for INFINITY ENGLISH ACADEMY — 600+ students & 1,200 parents; and an e-commerce system serving 200,000+ users. - Software Engineering student at PDP University on a 100% merit scholarship. Core stack: Python · FastAPI · Django/DRF · Aiogram 3 · OpenAI API · pgvector/HNSW (RAG) · PostgreSQL · Celery · Redis · Docker · CI/CD · nginx You get clean, documented code and a live, running system — with a README and a walkthrough, so you're never left with code that only works on my machine. If you want AI that reaches production, not a proof-of-concept that dies in staging, let's talk about your use case.