AI CIRCLE x ENCORD | VIRTUAL EXPERT AGI PANEL
Every few months someone declares AGI imminent, and the argument gets made almost entirely by people working on language models. This conversation asks the question differently.
Join Encord and AI Circle for a live virtual session with researchers and engineers building AI that has to work in the physical world, where a model cannot talk its way out of being wrong. Robots drop things. Perception systems miss edge cases. The gap between a benchmark score and a system you would trust is measured in broken hardware.
Moderated by Vineeth Velmurugan, Robot Learning Lead at Encord, with speakers from Amazon, NVIDIA and Cobot.
What we’ll get into
Whether the path to general capability runs through scale, or through data we do not yet have
What real-world training data actually looks like: egocentric video, teleoperation logs, EMG and brain-wave signals
Evaluating systems that act rather than answer, and why final-output scoring stops being enough
The long tail in perception, and what it tells us about generalization
Where human judgment stays in the loop, and whether that is a limitation or the design
What “general” means when the system shares physical space with a person
Who should join?
Researchers and engineers working on robotics, perception, evaluation, post-training, or the data pipelines underneath them. If you have watched a system clear a benchmark and then fail in deployment, you will have something to add.
Free to attend. Registration required.
Speakers
Moderator
Vineeth Velmurugan · Robot Learning Lead, Encord
Vineeth built Encord’s internal team producing real-world training data for physical AI, spanning egocentric video, teleoperation logs, EMG and brain-wave signals.
Rajat Kumar · Applied Science Manager & Science Lead, Amazon | Rajat leads science for robotics at Amazon, after earlier work on Alexa’s language understanding models and LLM research at Amazon AGI.
Siddharth Kothiyal · Senior Machine Learning Engineer, NVIDIA | Siddharth builds real-time deep learning and computer vision systems for autonomous vehicle perception at NVIDIA.
Tzu-Yun Hsu · Robotics Software Engineer, Cobot | Tzu-Yun works on human-robot interaction at Cobot, and was previously an AI Resident at 1X designing more natural interactions with the NEO humanoid.
About Encord
Encord is the data development platform for AI, enabling teams to manage, curate and annotate multimodal data across video, audio, images and text.
About AI Circle
AI Circle is an invite-only community of AI practitioners advancing the frontier, with chapters in San Francisco, New York, Seattle, Toronto, London, Paris, and campus circles at MIT and Stanford.