| |
|
| |
Building foundation models for the physical world, whether VLAs or world models, means wrangling massive video & sensor datasets, running large-scale training & simulation, & iterating fast.
The more data these models see, the more capable they get.
Join us for an evening of technical talks & networking with fellow ML & robotics engineers & builders to see how they are tackling this new wave of physical AI innovation.
AGENDA OVERVIEW
6:00 PM - Doors open, food, & networking
6:15 PM - Talks
Ury Zhilinsky, Founding Technical Member + Mouli Sivapurapu, Head of Data & App Eng, Mind Robotics
Nehil Jain, Forward Deployed Engineer, Anyscale
7:15 PM - Open networking
SESSION DETAILS
From Factory Floor to Training Run: What Works Today, What's Still Hard
Mind Robotics trains robot foundation models on egocentric video captured on live automotive production lines, plus teleop & third-party data, in a fast-growing corpus on a path to many petabytes. This talk walks through the pipeline end to end, from capture on the factory floor through QC & labeling to the training dataloader, covering what works in production today & the problems still ahead. Expect the patterns that worked & an honest list of what we haven't solved yet.
Speaker Bios:
Ury Zhilinsky leads software & AI at Mind Robotics, which makes AI-powered mobile robotic platforms for industrial manufacturing, with Rivian as its first manufacturing partner. Before Mind, his work spanned Waymo, Nuro, & Physical Intelligence.
Mouli Sivapurapu leads data & applications engineering at Mind Robotics, building the infrastructure that carries video & teleop data from the factory floor through QC & labeling into model training. Prior to Mind, he worked at Scale AI, Apple & Samsung.
Hosted by:
Anyscale, the company behind the Ray open source project, is a fully managed AI platform for foundation model-scale workloads. AI teams use it to run data, training, sims, inference, & RL workloads at scale on any cloud or on-prem.
|
|
|
|
|
|
|
|