Build a personalized Commute agent in Python with Hopsworks, LangGraph and LLM Function Calling
summary
A working Python pipeline combines Hopsworks, LangGraph and LLM function calling to power a personalized commute agent that ingests Stockholm public‑transport events from Kafka, computes 30‑minute, 1‑hour, 6‑hour windows aggregations, trains an XGBoost delay model, and exposes a LLaMA conversational layer that generates and runs Python to compute exact statistics. The system iterates on faulty generated code, logs results in feature groups, and serves real‑time predictions; LLaMA‑4 is considered too GPU‑heavy so LLaMA‑3x is used.
AI-written summary · about
chapters (13)
- 0:00Personalized commute agent intro
- 1:54Static LLM knowledge limits
- 3:20Retrieval augmented generation
- 5:53Dynamic context with function calls
- 8:02LangGraph workflows
- 10:38Accurate computation code
- 13:52Hopsworks FDI pipeline
- 15:19Demo architecture pipeline
- 21:40Deploying models
- 22:24LangGraph graph structure
- 23:49Live commute agent demo
- 25:15Handling code failures
- 26:18Q&A
related talks
38mBuild a personalized Commute virtual assistant in Python w/ Hopsworks & LLM Function Calling
47mBuilding Python based AI Systems with LLMs
01h 59mBuilding Agentic and Multi-Agent Systems with LangGraph
41mBuilding LLM Agents – Challenges and Learnings
09mBuilding LLM Agents with Tool Use
35mBuilding LLM-based Agents
01h 27mBlazing the AI Trail: Using LangGraph to Conquer the Oregon Trail
01h 43mBuilding LLM Agents (Hackathon)