I Built a Production AI Project Manager — LangGraph + NVIDIA NIM + Notion/Jira/Slack
In this video I walk through an AI Project Manager I built from scratch — a fully agentic system that autonomously plans, researches, and executes complex project tasks.
This isn't a tutorial project. It's production-grade AI engineering: real integrations, real observability, and patterns you can actually use at work.
What's covered:
00:00 — Introduction
00:30 — Architecture walkthrough (LangGraph, human-in-the-loop, NVIDIA NIM)
01:30 — Live demo (POST → SSE stream → report → approve → Notion/Jira/Slack)
03:00 — Code walkthrough (graph.py, nodes.py, main.py)
03:45 — Monitoring (Grafana + Prometheus + LangSmith)
04:15 — Key takeaways
Tech stack:
LangGraph · LangChain · FastAPI · NVIDIA NIM · Ollama · Tavily · Notion API · Jira API · Slack SDK · Prometheus · Grafana · LangSmith · Docker · Python 3.11
Get the full source code:
👉 https://bzddbz.gumroad.com/l/project-...
Technical contents:
👉 https://bencedesvari.substack.com/
Other projects:
👉 https://bzddbz.eu
📌 If you found this useful, like and subscribe — I post about production AI engineering, agentic systems, and Python.