Podcast
Using Knowledge Graphs & LLMs for Automotive R&D: RAG, Graph ML & Crash Simulation
Open original DataTalks.Club episode
Using Knowledge Graphs & LLMs for Automotive R&D: RAG, Graph ML & Crash Simulation
Original Episode
Use these links for the canonical episode and media sources.
- Open the original DataTalks.Club podcast page
- Watch on YouTube
- Listen on Spotify
- Listen on Apple Podcasts
Episode Overview
How can knowledge graphs and large language models (LLMs) be combined to accelerate automotive R&D — from crash simulation insights to reproducible reports? In this episode Anahita Pakiman, a data scientist-engineer who moved from mechanical engineering and finite element analysis (FEA) into applied AI and now works as Senior Knowledge Graph-Data Scientist Consultant at brox IT-Solutions, walks through practical strategies and tradeoffs.
People
Use these links to connect the episode to guest notes.
Chapter Summary
Use these checkpoints to decide whether to open the source transcript.
- 0:00 - Episode Introduction
- 1:40 - Guest Bio: career path from mechanical engineering to applied AI
- 2:57 - Guest Background & Career Transition
- 5:37 - Applied Mechanics & Finite Element Analysis (FEA) overview
- 8:05 - FEA vs Machine Learning: numerical modeling vs data-driven approaches
- 8:50 - Optimization, Topology & Semantic Reporting in crash simulations
- 15:58 - Knowledge Graphs for Automotive R&D: motivation and Neo4j adoption
- 20:32 - Graph vs Tabular Representations: visualization, clustering, load-path detection
- 26:15 - From Knowledge Graphs to Computational Graphs (NetworkX & graph analytics)
- 28:00 - Graph Data Science & Graph ML: similarity measures and SimRank
- 33:43 - Combining Knowledge Graphs & LLMs: grounding and retrieval-augmented generation
- 38:10 - Text Chunking, Embeddings & Vector Databases vs Knowledge Graph Semantics
- 39:56 - Prompt Templates & KG-driven Retrieval (Cypher-based examples)
- 40:23 - RAG vs Transfer Learning: embeddings, fine-tuning, and distinctions
- 42:42 - Trust, Hallucination & Verification Limits of LLM-extracted Knowledge
- 44:13 - ADPT-LRN-PHYS Project Overview: LLM + KG for adaptive learning and paper
- 47:10 - Paper Parsing & KG Visualization: sections, keywords, PageRank and reference
- 54:18 - Project Challenges: automating graph generation and scoping the demo
- 55:36 - Deployment & Frontend Issues: Streamlit limits and state management for graph
- 57:46 - Learning Resources: graph ML courses, Jure Leskovec, Graph Conference and