The industry treats safety and efficiency as a trade-off. It doesn’t have to be.

Meet the team
Our Story

The Gap

After building the models that taught autonomous vehicles how to keep a safe distance from people and objects in motion, our founder, Denis, saw the same intelligence missing somewhere else entirely: industrial robotics.

Robots freeze the moment they detect a person or object nearby, treating the stop as the only safety mechanism they know. But every stop means lost time and productivity.

The industry’s fix was to slow down, or fence people out. Neither one solved the actual problem.

Industrial robot working alongside people

The Birth of scaledrive.ai

Denis brought together a small team with the exact background this problem needed: deep experience in autonomous driving, large-model research for robotics and humanoids, and provable safety focused on safe human-robot interaction.
‍
Together they built the full autonomy stack the industry was missing: prediction fast enough to keep robots moving, with safety provable enough to certify.

Our goal is to bring this technology into ports, shipyards, and industrial sites, building toward a future where robots and humans work side by side.

scaledrive.ai team and robots on an industrial site

Where We Are Today

Our autonomy stack is deployed in the real world, at a major UK port and with a European heavy industry manufacturer, across both AMRs and humanoids.

TÜV and the Autoware Foundation have both signed letters supporting our approach to human behavior prediction and certification. We're backed by a16z speedrun, Forum Ventures, and Unpopular Ventures, part of NVIDIA's Inception Program, and a member of the Autoware Foundation.

scaledrive.ai intelligence layer deployed on site
Our Team

Built by experts in safety and autonomy

Denis Azarov

Denis Azarov

Co-founder · CEO

Engineering and innovation leadership at BMW; Prior startups reached thousands of users and worked with leading enterprises in transportation and autonomous driving sectors.

Matthias Meyer

Jakob Thumm

Co-founder · Head of Research

TU Munich PhD on provable safety for robotics, Stanford postdoc. Author of the provably safe RL work our safety layer is built on.

Tim Salzmann

Tim Salzmann

Founding ML Researcher

PhD researcher behind Trajectron++, benchmark work in large models for robotics at Stanford and Google DeepMind, autonomous driving at BMW.

Smiling woman with long hair and glasses wearing a patterned top with hand near her head.

Yutong Xin

Safety & Robotics Engineer

Safety-layer development at TU Munich, robotics AI software and simulation at Fraunhofer.

Yutong Xin

Ott Männik

VP of Engineering

Led development and launch of Autoware-based autonomous shuttles deployed across 8+ use cases in Europe and Japan.

Jakob Thumm

Ingmar Sell

Senior Autonomy Engineer

Senior software engineer for autonomous shuttles deployed in more than eight use cases around the world.

Our Advisors

Advised by the best in the industry

John Hanke

Investor

VP Google GEO (led inception of Google Maps) Founder Niantic (Pokemon Go) & Niantic Spatial (spatial intelligence for robots).

Yevgen Chebotar

Yevgen Chebotar

Shareholder · Robotics AI

Early contributor and shareholder. The leading contributor to the concept of VLAs, now in World Model Research at Nvidia. Led AI VLA efforts at Figure; formerly Google DeepMind.

Black and white portrait of a smiling middle-aged man with short hair and a beard.

Matthias Meyer

Investor & Advisor

Corporate innovator; co-founded BMW’s Venture Client unit (BMW Startup Garage); PhD in Safety Engineering & ML. Led the development and deployment of Europe's first large-scale end-to-end interceptor drone systems.

FAQ

Frequently Asked Questions

We build the full end-to-end autonomy stack, differentiated by advanced prediction and provable safety. Think of it as the reflexes a good driver has in traffic, built into the robot instead.

Stopping is the current default, and it kills productivity. Every stop compounds across a shift, a yard, a fleet. Our layer lets the robot read the person’s movement and route around them instead, so the work never pauses.
AMRs, humanoids, and quadrupeds. The underlying prediction model is the same across all three. Only the physical task changes.
Yes. The safety layer is architected against existing standards like ISO 3691-4, and we’ve engaged TÜV early in the process to validate the approach.
We’re hardware-agnostic. The software runs on ROS, the industry standard, so it deploys onto AMRs and humanoids you already operate or plan to bring in.
Today, we’re operating in ports and shipyards. Soon, we’re expanding to airports, oil refineries, and robotics companies looking to license our safety layer directly.

Where would robots work if they could do it safely?

Tell us where automation stalled. We’ll take it from there.

Talk to us