Robots that learn on the job
TL;DR: We deploy robots that learn new tasks from a single human demonstration in the factory. A new task shows up on the floor? You just film it once on your phone and the robot picks it up. No sitting around collecting data. No engineer flying out to your site. No overnight fine-tuning. You show it, and the robot learns.
❌ The problem
Manufacturing is full of constantly changing, high-mix production tasks, which are among the hardest to automate. A line that's packing one product today is building something completely different next week. Every factory owner we talk to says the same two things: they can't find enough people to do the work, and the automation solutions they've tried have been too expensive and too rigid, often ending up gathering dust in a corner.
✅ What we do
We're developing a robot foundation model whose core capability is adapting instantly to new variations of a task. You give the model a human demonstration video as context, and it bridges the embodiment gap to execute the task itself. This in-context learning is the missing piece in today's VLA approaches, and it's what changing manufacturing environments need.
We deliver the complete system: robot hardware, foundation model, and on-site deployment. We're already deploying robots at customer sites.
👋 Who we are
Between the two of us, we have 20+ years of experience in robotics. Navid was part of the core team that productized Spot at Boston Dynamics and led manipulation projects there; before that, he was at Rethink Robotics. Mojtaba led whole-body control and bimanual manipulation efforts at Amazon Robotics and was a Research Assistant Professor at Northwestern University.
🙏 Our ask
Do you have a manufacturing task that needs automating, or know someone who does? Let us know and we'll automate it.
Reach us at: founders@shiraz.ai
