Agenda

September 23
8:00am
Doors open
Registration + Light Breakfast
9:00am
Opening remarks
Welcome to Actuate 2025
9:20am
Keynote
The Physical Turing Test: Solving General Purpose Robotics
The age of generalist AI agents for the physical world is here. Join Jim Fan, Senior Research Scientist and co-lead of NVIDIA GEAR Research Lab, as he outlines the major challenges facing the “robot revolution” and how the community is overcoming them with techniques in simulation, synthetic data and accelerated computing.
9:50am
Presentation
Solving Real-World Navigation for Humanoid Robots in Dynamic Environments
Vivian Chu, Co-Founder & Chief Innovation Officer at Diligent Robotics, will discuss technical strategies for physical interaction with legacy hospital infrastructure—such as navigating elevator buttons, automated doors, and access-controlled zones—and walk through key system design decisions made to enable reliable, respectful, and safe humanoid robot navigation in busy, human-centric indoor environments.

As hospitals represent some of the most complex, safety-critical environments for robots, this talk will highlight how recent advances in applied AI are pushing the boundaries of what robots can achieve in real-world collaborative scenarios.
10:10am
Conversation
Fireside Chat with Kyle Vogt
Kyle Vogt is the Founder and CEO of The Bot Company, a robotics startup making affordable household robots. He previously founded Cruise, which was sold to GM for $1 billion and ultimately completed 250,000 driverless rides. Previously, Vogt was a cofounder of Twitch. He studied Computer Science and Electrical Engineering at MIT.
10:40am
Break
Stretch & Connect
11:00am
Presentation
Foundation Reward Models for Robot Learning
Jason Ma is co-founder at Dyna Robotics. He received his PhD from University of Pennsylvania, where he studied scalable reinforcement learning algorithms and foundation models for robot learning. His research has been recognized with honors such as RSS Pioneers, Apple and OpenAI Fellowship, Best Paper Finalist at ICRA 2024, Top 10 NVIDIA Research Projects of 2023, and covered by popular media such as the Economist, Fox, Yahoo, and TechCrunch.
11:20am
Presentation
End to End Robotics in Rust
Vivek Bagaria is the Autonomy Lead at Matic Robots, a company with the mission of automating home chores to give back people their time and energy.Vivek holds a PhD in Electrical Engineering from Stanford University and a Bachelor's from IIT Madras.At Matic, he introduced Rust and spearheaded its adoption across the whole software stack — establishing robust technical foundations throughout the company.
11:40am
Presentation
Building with Nix for Embedded Systems
Embedded systems often require customization throughout all levels of the firmware/software stack (custom toolchains, platform firmware, kernel drivers, device trees, system libraries, and applications), which can quickly become a unmanageable tower of complexity for the build system.

Nix does a particularly good job of allowing a developer to comprehensively define, track (with version-control), and reproducibly build system software with these customizations in a way that systematically prevents many of the build problems associated with using traditional tools.

In this presentation, I'll share some of the wins and lessons learned as a result of adopting and heavily relying on NixOS.
12:00pm
Presentation
Unchaining Physical AI
The advancement of Physical AI is fundamentally crippled, not by the algorithms alone, but by the friction between the AI, the robot, and the physical world it operates in. To build for sustained, real-world operations  , we must conquer the entire stack. This talk presents a holistic approach that simultaneously optimizes the AI, its physical embodiment (the robot), and its operating environment.
 
We've learned that AI performance is acutely sensitive to physical reality. By understanding this interplay, we can strategically improve robot hardware and environment to achieve maximum gains in AI capability. We introduce IRIS, the MCP-of physical AI, capable of letting AI’s understand and command the maximum out of HW systems.
 
This architecture enables a new evolution of intelligence. We move beyond simple "Dextrous AI" for manipulation—recognizing that, like the best LLMs can't play Chess well, just fine motor control isn't enough. Doing a job professionally, and well - requires deep domain understanding of the physical labor and how to do the job  with high competence, with professional experience guiding your day to day plans  . This reframes the debate around form factor; instead of merely recreating ourselves in humanoid form, we must ask: if we are the architects of this new workforce, what is the optimal form for the complex jobs we need done?
12:30pm
Break
Lunch
2:00pm
Panel
Middleware vs. Minimalism: Should you Build on ROS or Start from Scratch?
Robotics teams wrestle with whether to adopt ROS, extend it, or rip it out entirely. Some swear by its ecosystem; others view it as tech debt from day one.
2:40pm
Presentation
Multi-Robot Teaming to Serve Food
A common human task in food assembly is having multiple workers place the same ingredient into alternating bowls moving at a high rate of speed to increase throughput. While this seems simple at first, detecting which bowl already has an ingredient in it with 99.999% accuracy over thousands of ingredients proves challenging for robots.

In this presentation I'll share how we get robots to achieve similar performance to humans, deep dive into the various engineering efforts we've taken to solve this challenge in production (classical computer vision, mesh networking, SLAM, Siamese neural networks, KNN) and how our robots have served millions of bowls utilizing the features we built.
3:00pm
Presentation
Last 100 Meters: Autonomy Challenges in Scaling Drone Delivery to Every Home
With over 100 million autonomous miles flown across eight countries and four continents, Zipline is transforming access t o healthcare, consumer products, and food. In this talk, we’ll explore how Zipline designed Platform 2 for safe, precise, and magical deliveries at scale. We’ll dive into the autonomy challenges of navigating the incredible diversity of real-world homes and delivering packages exactly where customers want them—threading through tree cover in wind or dodging powerlines at night. We’ll share how data and simulation fuel our ML-driven autonomy stack and highlight key challenges ahead.
3:20pm
Break
Stretch & Connect
3:40pm
Conversation
Co-Piloting Autonomy: Engineers and Operators Scaling Real-World Robotics Together
Deploying autonomous robots in the real world is a complex challenge that goes beyond perfecting algorithms. Unpredictable environments, safety requirements, and operational complexities can derail even the smartest robots once they leave the lab. This fireside chat explores how combining advanced autonomy with human expertise helps bridge that gap. By effectively "co-piloting"—pairing engineers and field operators throughout the process—teams can scale robotic systems in dynamic real-world settings more safely and efficiently.

Heidi and Jess will dive into how to build robust feedback loops from field operations back into development. They will share practical methods for observing and measuring robot performance in real time, defining reliability metrics, and handling edge cases encountered on the job. The session also explores strategies to scale an autonomous fleet without compromising visibility or safety, and how clear handoff logic ensures smooth transitions between autonomy and human control. Attendees will gain insight into effective collaboration patterns between engineering and pilot teams that improve system reliability and ensure deployments can scale successfully.
4:10pm
Presentation
Accelerating Autonomy Development with the Hivemind SDK
Shield AI has always been driven by a singular mission: protecting service members and civilians with intelligent systems. We initially built our core autonomy technology, Hivemind, for indoor exploration on a small quadcopter, but then extended it to increasingly complex aerial systems. Recently, Shield AI reached an exciting new milestone: we distilled years of expertise fielding autonomy on platforms into the Hivemind SDK, an ecosystem offering production-ready autonomy middleware, robust autonomy libraries, and tools for fast and rigorous autonomy development cycles.
This talk will highlight case studies addressing challenges in autonomy development, showcasing how we leverage Hivemind Forge, including Foxglove, to rapidly iterate through the fly-fix-fly cycle and field resilient systems. First, we walk through how we set up the autonomy problem and use a diverse set of visualizations to debug difficult failure modes. Next, we describe our approach to airworthiness and certifiability, using key visualizations to monitor and analyze the performance of our multi-modal runtime assurance framework. Fielding autonomy quickly to large, expensive hardware requires high levels of assurance – possible only with appropriate safety wrappers. Finally, we demonstrate how we approached development and validation of heterogeneous teaming in communication-constrained environments by having agents balance the search task while maintaining the communications bridge for operator insights.
4:30pm
Presentation
Fast-Loop Autonomy: How Observability at Scale Accelerates Robotics Development
As Slip Robotics scales fleets of SlipBots for real-world dock operations, the challenge shifts from making autonomy work once to making it work everywhere—safely, efficiently, and consistently. This talk unpacks the architecture that makes that possible. Dennis will show how Slip uses Foxglove, custom data pipelines, and internal monitoring tools to:

• Surface meaningful trends from across the fleet
• Link autonomy failures to real-world conditions
• Automate debugging and regression detection
• Enable rapid iteration on autonomy software, even with tight customer uptime requirements"
5:00pm to 6:30pm
Day 1 - Reception
Happy Hour Networking
September 24
8:30am
Doors open
Light Breakfast
9:00am
Keynote
Robotic Foundation Models
General-purpose models, such as LLMs, have transformed how we think about AI: instead of specialized systems that are designed specifically to solve one task really well (e.g., AlphaGo), general-purpose foundation models provide for a generalist solution that can utilize data from many tasks, and solve a wide variety of problems based on user requests. What would this look like in the world of robotics? At Physical Intelligence, we are developing such generalist robotic foundation models, with the aim of enabling one model to control any robot to perform any task. In this presentation, we'll discuss some of our recent advances that could make this possible, including advanced second generation vision-language-action (VLA) models, methods for integrating complex hierarchical problem-solving, and training techniques for facilitating knowledge transfer from web-scale pretraining.
9:30am
Keynote
The Future of Intelligent Automated Logistics
How do you orchestrate a fleet of thousands of robots in real-time to manage the movement of millions of cartons through a warehouse per day? In this session, James Kuffner, Chief Technology Officer of Symbotic – a pioneering leader in robotics – will discuss his transition to Symbotic, a company delivering advanced AI-powered logistics robot fleets to the global supply chain. The recent rapid progress in machine learning, computer vision and perception, as well as increased computing power have enabled a new generation of intelligent robots that can operate reliably in semi-structured environments like warehouses. Innovative new algorithms for multi-robot motion and task planning will scale the technology even further, creating exciting new opportunities for the future of intelligent automated logistics.
10:15am
Conversation
Fireside Chat with RJ Scaringe
RJ Scaringe founded Rivian in 2009 to help accelerate the world’s shift toward carbon neutral energy and transportation. Since starting Rivian from a clean sheet, he’s led the company to establish its vertically integrated technology platforms, build multi-program manufacturing capabilities, develop meaningful partnerships and set forth a strategic vision that transcends the profound change we’ll see within the transportation industry over the next few decades.
11:00am
Break
Stretch & Connect
11:30am
Presentation
From Pixels to Parts: Deploying Direct CAD-to-Robot Workflows in Construction
Reframe Systems is transforming modular construction of climate-resilient homes through software-driven robotic automation. Our platform enables direct-to-robot workflows that bring CAD models to life, allowing for highly customizable manufacturing at production scale. We will explore how we've deployed a vision-driven robotic end-point that seamlessly operates from our Onshape CAD environment. This direct connection enables automated generation and validation of robot instructions without manual translation layers. Our first robotic platform has built over 1000 linear feet of residential walls for customer projects, a milestone made possible by our emphasis on short-cutting development to intercept production as early as possible, rapid hardware iteration ( through our patent-pending universal magnetic fixturing among other design decisions), and vision-based adaptability. This foundation puts us on track to achieve 90% of all panels per project by the end of the year. Our approach highlights flexibility and scalability, empowering robotics workcells to dynamically adapt to diverse production demands.
11:50am
Presentation
Eyes for Humanoids and Advanced Robotic Applications Enabled by Marrying Optical Physics with ML Techniques
Robotics has gained significant momentum with the advancement of VLAs. Relying on traditional RGB cameras for vision is an affordable and scalable option, but they have limited capabilities in scenarios encountered by humanoids, wheeled indoor robots and manipulations. Even expensive lidars struggle to provide accurate and reliable depth in these scenarios such as detection of glass cups, glass doors, reflective indoor floors, shiny metal objects, thin wires, dark objects, etc. By using plenoptic sensing technology, we are able to provide significant benefits over contemporary vision systems thereby offering a good balance between cost and performance. In summary, we are able to provide the high quality of lidar at the cost of traditional RGB cameras. Learn how combining optical physics with transformers and other ML techniques is enabling sharper and scalable vision systems for the next wave of robotics. The technology has found product market fit and is currently in the phase of scaling deployments in applications such as delivery robots, robotic arms with end-effectors, indoor robots/humanoids, and autonomous vehicles.
12:10pm
Presentation
Lessons from Deploying 2,000 Delivery Robots Across Geographies
Serve Robotics is scaling to deploy 2,000 autonomous delivery robots across diverse geographies in 2025. In this talk, we’ll share our journey of building scalable autonomy for complex, unstructured urban environments. Central to our approach is a rapid deployment cycle that enables real-world learning, driven by a powerful fleet data flywheel. We’ll delve into the unique autonomy challenges of last-mile delivery, our high-level AI architecture, and the major technical initiatives for a robust and scalable autonomy stack.
12:30pm
Break
Lunch
2:00pm
Panel
Structure or Scale: How Much Should We Engineer into Robot Learning?
As E2E vision-language-action models gain traction, the robotics world is splitting into two camps: those who want to hand off everything to foundation models, and those who who believe in finely tuned, modular components.
2:40pm
Keynote
Foxglove
TBA...
3:00pm
Presentation
Unnatural Selection: Evolving Robotic Systems within Systems of Compounding Complexity
TBA...
3:20pm
Break
Stretch & Connect
3:40am
Presentation
Skild AI
Deepak Pathak is CEO & Co-Founder of Skild AI and Professor at Carnegie Mellon University working on developing an AI foundation model for robotics. He is known for pioneering contributions to AI including self-supervised learning, artificial curiosity and sim2real adaptation for robot learning. Deepak has received several prestigious honors including Sloan Fellow, MIT TR 35under35, Okawa Fellow, IIT Kanpur Young Alumnus Award and multiple Best Paper Awards. Deepak's research has been featured in popular press outlets, including The Economist, The Wall Street Journal, Forbes, Quanta Magazine, Washington Post, CNET, Wired, and MIT Technology Review among others. Earlier, he received his Ph.D. from UC Berkeley and his Bachelor's from IIT Kanpur with a Gold Medal in Computer Science.
4:00pm
Presentation
Advanced Autonomy for the Built World
Abstract: Bedrock Robotics is on a mission to transform construction by bringing full autonomy to specialized heavy equipment. In an industry where productivity has barely improved in decades, Bedrock is working to deliver safer, faster, and more efficient operations in challenging worksite conditions. This talk will share how many of the team's learnings from driverless cars and trucks at Waymo have applied to these complex machines and will highlight the development journey and approach toward a fully automated construction future.
4:20pm
Keynote
Frontiers in Embodied AI
Autonomous driving will be its first widespread deployment of the next wave of AI, Embodied AI.

In this talk, Jamie Shotton, Chief Scientist at Wayve, will explore the company’s Embodied AI technology and their recent global expansion. The talk will unpack Wayve’s AV2.0 approach, which uses a single end-to-end neural network to perceive, predict, and plan directly from data, and explain how that same model now drives—with little to no prior training—across the US, Canada, Germany, Europe, and Japan.

He’ll dive into the foundation model, safety-by-design architecture, and how data from diverse markets fuels rapid generalization without HD maps or costly sensors.
4:50pm
Closing Remarks
Thank you!
5:00pm
Day 2 - Reception
Afterparty Hosted by Rivian
Sponsors

Thank you to our conference sponsors.

Contact us to sponsor Actuate.
Platinum
Gold
eclipse ventures logoLightwheel logoouster logosegments.ai logo
Silver
tangram vision logo
Bronze
pathbreaker ventures logo

Tickets are limited. Get yours today.

Register