The multi-agent + RAG combo is what I'm living in right now. At AmberFlux EdgeAI I'm building SHEELA — an intelligence layer that crawls, scores, and runs controlled self-improvement loops on top of a FastAPI/PostgreSQL infrastructure backbone I architected from scratch. LangGraph for the agent orchestration, async pipelines throughout. But the project that feels most relevant to what you're describing is something I built earlier — a Persona RAG system that fused Karpathy, LeCun, and Mollick's thinking patterns from their writing and talks into a Telegram bot that responded in their synthesized voice. The whole point was: how do you get a specific person's reasoning logic out of their head and into a system. Sound familiar? Also built RepoGuard — a LangGraph + Groq Llama 3.3 70B agent that reads GitHub PRs and issues and surfaces intelligence from them. And shipped Repoverse, a full Next.js + iOS app to the App Store in 175 countries. On the client side — I run Pixel Elevate, an AI automation agency where I've built and delivered systems for clients across the US, UK, and Canada. So translating a non-technical founder's vision into a shipped product, managing expectations, and owning the outcome end-to-end isn't new territory for me. I'm a final-year engineer graduating May 2026, so I won't pretend I've managed engineers. But I've managed clients, shipped production systems solo that real users depend on, and I use Aider + Claude Code + Cursor as daily drivers — not as novelties. The Persona RAG bot feels like the most honest answer to your "system you're proud of" question. Happy to walk through it. Filling out the interest form now. GitHub: https://github.com/mandanajignesh-byte | HuggingFace: https://huggingface.co/Jignesh2619
LinkedIn:www.linkedin.com/in/jignesh-mandana-936405349