You Can’t Rerun Sunlight: Designing ML Data Architectures for Physical AI
What this session covers
Large language models are transforming how we build software, but physical AI systems expose a hard limit: you cannot recompute reality.
When robots, sensors, and production systems interact with the real world, failures are causal and time-based, not semantic. You cannot go back to record sunlight you missed, an observation you never captured, or robot telemetry lost to fragile connectivity. Backfills can repair a dataset while changing our understanding of what the system actually knew at the time.
In this talk, I will show how these constraints change how we design ML data platforms for robotics, agriculture, manufacturing, and other physical-world domains. We will look at how to capture irreversible data reliably at the site where it happens, preserve the difference between observed and reconstructed data, handle late and out-of-order events without rewriting history, and build a shared historical data backbone for analytics, training, and debugging.
I will close with where LLMs and agents fit into this workflow, and why they can reason about evidence but cannot recover evidence that was never captured.
World Congress 2026 North America
Related talks
- From Simulation to Reality: Overcoming the Data Challenge in Physical AI
Mitesh Patel · NVIDIA Corporation
Fri 25 Sep · Stage 1
- Edge AI: Running Agentic Intelligence Where Internet Can't Reach
Nitin Eusebius · AWS
Thu 24 Sep · Stage 5
- No Single Model to Rule Them All: Building Resilient AI Agents Across Open & Closed LLMs
Emmanuel Acheampong · Crusoe AI
Thu 24 Sep · Stage 6
- Anatomy of an AI Request: Where Latency and Cost Are Really Born
Dan Fu · Together AI
Thu 24 Sep · Stage 1