Master enterprise-scale generative AI
summary
ReconAI runs in production classifying rail faults with a RAG + agentic workflow (S3, Glue, Athena, EC2, Bedrock/Anthropic Claude) and achieves 96% classification accuracy from an initial 80% in PoC, covering 500+ fault classes. The system logs prompt and model experiments (Bedrock prompt management, FMEval, Raggus, LLM‑meter), shows full transparent reasoning for each classification, prioritizes accuracy over latency, and reduced manual engineering time enough to drive prioritised maintenance projects and continuous iterative improvements.
AI-written summary · about
chapters (9)
- 0:00Introduction and host welcome
- 1:55Path to production for GenAI
- 4:49Experiment tracking and evaluation
- 10:28Prompt management importance
- 16:07Monitoring GenAI workloads
- 17:12Data quality and ground truth
- 18:36Rio Tinto’s case study
- 24:18Journey from proof of concept
- 31:37Key takeaways and closing
related talks
32mGenerative AI for Observability: Automating Root Cause Analysis in Modern IT Systems
13mDeploying Enterprise-Grade Compliant Generative AI
24mScalable GenAI in Production
38mBuilding AI responsibly with a GRC strategy
40mRolling out Enterprise AI: Tools, Insights, & Team Empowerment
44mEnhance Your Generative AI Applications with RAG
41mGenerative AI: bliss or miss?
38mGenerative AI - Architectures and applications in depth