Backed by Y Combinator

Robots that learn on the job.

Shiraz builds a robot foundation model that learns a new manufacturing task from a single human demonstration — delivered working on your line: hardware, model, and deployment.

Watch the demo Get in touch

San Francisco, CA · 37.7749° N, 122.4194° W · YC S26

Sec. 01 Field demonstration

Watch the demo

Still frame from the Shiraz AI launch demo: a robot arm working at a manufacturing station
Fig. 01 Shiraz AI — YC S26 launch demo (2 min)

Sec. 02 Operating principle

What we do

  1. 01Input

    Show

    A person demonstrates the task once, on video. No programming, no CAD, no teach pendant.

  2. 02Adaptation

    Learn

    Our foundation model adapts in context from that single demonstration — no retraining, no data-collection campaign.

  3. 03Deployment

    Work

    The robot runs the task on your line. We deliver the full stack: hardware, model, and on-site deployment.

Note Why now

High-mix manufacturing changes too fast for traditional automation. Factory owners tell us the same two things: they can’t hire enough people, and the automation they’ve bought is too rigid and too expensive — so it sits idle. We built Shiraz for lines that change every week.

Status — operational

Our robots are already operational at customer sites.

Sec. 03 Personnel

Team

Two decades of combined experience shipping real robots.

Navid Aghasadeghi

Co-founder & CEO

Part of the team that commercialized Spot at Boston Dynamics, where he led manipulation projects. Previously Rethink Robotics.

Mojtaba Mozaffar

Co-founder & CTO

Led whole-body control and bimanual manipulation at Amazon Robotics. Formerly Research Assistant Professor at Northwestern University.

Experience from
Boston Dynamics Amazon Robotics
Backed by Y Combinator

Sec. 04 Open positions

We’re hiring world-class roboticists.

Foundation models, manipulation, and real-world deployment. San Francisco, CA.

founders@shiraz.ai

Direct to the founders — no recruiters, no forms.