Robot Learning #03 — Simulation vs Reality & Building the Robot Data Engine

Simulation can make experiments repeatable and help teams explore scenarios that are difficult to collect in the physical world. Real-world data brings its own variation, constraints, and surprises. Building a useful robot data engine means understanding what each source contributes and how to learn from the gap between them. Join HackerSquad and Bright Data for the third Robot Learning meetup, focused on simulation versus reality and the data workflows behind physical AI. We'll discuss how teams collect, curate, and evaluate data for VLAs and world models, and how deployment failures can inform the next training cycle. We'll begin with a few short demos, continue with a practitioner panel on the tradeoffs behind these systems, and close with networking for the researchers, founders, and engineers building them. What we'll explore Combining simulated, real-world, and demonstration data for robot learning. Understanding transfer gaps and choosing evaluations that reflect the tasks a robot needs to perform. Building a repeatable loop from data collection and curation to training, evaluation, and feedback from deployment. Who this is for Robotics founders, ML engineers, researchers, simulation developers, and data teams working on VLAs, world models, and physical AI. Bring a question about your training pipeline, an evaluation challenge, or a lesson from moving into the real world. Schedule All times are Pacific Time (America/Los_Angeles).