Robots to package billions of e-commerce orders in warehouses
Ultra builds AI-powered robots that automate the dull, repetitive labor still done by people in American warehouses. We’re starting with e-commerce order packaging in fulfillment centers — where a worker puts items in a box, seals it, and labels it. Traditional automation isn't working for warehouses because it's costly, rigid, and often underutilized. Ultra’s robots are different: they’re easy to deploy, adaptable, and powered by AI that’s trained through examples.
Active Founders
Jon Miller Schwartz, Founder
Building intelligent-industrial robots at Ultra. Prev Arena AI, Voodoo Manufacturing (YC W17), Body Labs, and Layer By Layer (YC S13).
Max Friefeld, Founder
Max is the co-founder and COO of Ultra. He studied Computer and Electrical Engineering at Harvey Mudd College ('13), and founded two YC-backed companies, Layer By Layer (S13) and Voodoo Manufacturing (W17). Immediately prior to starting Ultra, Max led teams at Boston Consulting Group optimizing operations for the largest logistics networks in the world and developing go-to-market strategy for private equity backed robotics companies.
Oliver Ortlieb, Founder
Oliver is the co-founder and CTO of Ultra. Before co-founding Ultra, Oliver Ortlieb was the Co-founder & CTO at Voodoo Manufacturing where he built the software and systems to support one of the largest 3D printing factories in the world. Prior to that, he co-founded Layer By Layer, a secure marketplace for 3D-printable products. He holds multiple patents in the field of 3D printing, and received his BS in Computer Science from Harvey Mudd College.
Chetan Parthiban, Founder
Chetan is a Co-Founder and Chief Scientist of Ultra. He studied Mathematics and Robotics at the University of Pennsylvania, where he received both his BA and MSE. Chetan has deep experience in applied AI as he was one of the first machine learning scientists at Arena, where he focused on applying machine learning to solve high frequency decision making problems.
Company Launches
TL;DR: Ultra robots automate e-commerce order packaging and returns in fulfillment centers. Traditional automation isn't working for warehouses because it's costly, rigid, and often underutilized. Our robots are different: they’re easy to deploy, resilient to changing environments, and powered by AI that’s trained with examples.
Hi everyone - we’re Oliver, Chetan, Jon, and Max. After years of building and scaling manufacturing and robotics companies (Layer By Layer, S13, and Voodoo Manufacturing, W17), we’re now focused on bringing automation to the industries that need it most.
We believe we’re at a moment in history when robots can be made accessible and capable enough for mass adoption - and we need it more than ever.
Labor is harder to come by than it’s been in decades.
Imagine running a warehouse: you hire 20 temp workers to come in one day, but only 15 show up in the morning, and just 5 return after lunch. That’s the reality one of our partners told us they face.
In 2018, the U.S. entered its first labor shortage in over 60 years. Today, we are short over 1.3 million workers, causing companies to go understaffed or rely on high-churn temp labor.
And yet, traditional warehouse automation isn’t being adopted fast enough.
🤖 Ultrabots
Ultra is building robots that can be dropped in at existing workstations and quickly trained to do repetitive tasks. They are more cost-effective than human labor and can operate around the clock, giving your team ✨superpowers✨.
We use reliable off-the-shelf hardware so we can focus on rapid deployment. The key to unlocking the next million robots is data scale, and the fastest way to get there is to put robots in real-world environments doing useful tasks.
⏰ Why Now?
Recent AI breakthroughs mean we can control robots with neural nets trained with examples, rather than with explicitly programmed routines.
Above is an example of a fully autonomous picking policy we trained on ~2hrs of data from teleoperating our research arms. The two RGB camera feeds are the input to the model, which outputs the joint+gripper positions at 10Hz.
Examples of emergent behavior learned from the training data, but not explicitly programmed like what would be required in traditional automation.
🙏 Asks
We’d love to connect to e-commerce 3PLs and large brands that handle their own fulfillment. You can reach us at founders@ultra.tech. Thanks!