#movingpeople-meets Foretellix

It’s not every week that a company becomes a key part of NVIDIA’s autonomous driving platform. But that’s exactly what happened when Foretellix announced its new integration with NVIDIA - integrating the company’s Foretify Physical AI toolchain with NVIDIA’s DRIVE AV platform. This collaboration connects Foretellix’s scenario-based safety and synthetic data generation capabilities with NVIDIA’s perception, simulation, and autonomous driving technologies, shaping how autonomous vehicles are trained, tested, and verified.

To explore what this means for the future of autonomy, I sat down with Ziv Binyamini, CEO and co-founder of Foretellix.

#movingpeople is a part of Mobility Business  - a consultancy dedicated to "All Things Mobility", focused on growth.

Let’s start with the news - Foretellix and NVIDIA. What exactly does this news mean?

This is an expansion of our ongoing relationship with NVIDIA. We’re integrating our Foretify Physical AI platform directly into NVIDIA DRIVE AV, which allows developers to train, test, and validate their AI drivers using our tools inside NVIDIA’s simulation environment. It connects our scenario-based safety and synthetic data tools with NVIDIA’s perception-level simulation. Now developers can build, train, and verify autonomous systems with our measurable safety.

The focus here is on Physical AI - where AI doesn’t just analyze, but acts. Safety will define its future, and we’re making sure that safety is measurable and actionable.

Now a step back for readers less familiar with industry terms and with Foretellix, what’s the core problem you solve?

Foretellix exists to make autonomy safe - extremely safe. Building something that drives is easy. Making it safe, including all the edge cases you didn’t anticipate, is the real challenge. Autonomous vehicles are heavy, fast, and interact with humans. The smallest mistake can cause enormous damage. That’s why the biggest pain in the autonomous industry is proving safety - efficiently.

You could, in theory, test a million cars for ten years and spend billions to validate safety. But that’s not a good use of time and money, not to mention that most companies don’t have that money to begin with. Foretellix provides the software tools that let companies test millions of virtual scenarios - safely, quickly, and cost-effectively.

Physical AI. You use the term, and I’ve been hearing it increasingly lately. Walk us through that and what it means for Foretellix.

We’re at the frontier of AI - Physical AI - where intelligence interacts with the physical world. Think of cars, robots, drones. These systems perceive the environment, reason about it, and then act. Our tools make sure that action is safe.

Until now, AI in autonomy was mainly used for perception - recognizing pedestrians, traffic lights, or lane markings. But now, we see end-to-end AI; neural networks that take sensor inputs and make full driving decisions. It’s powerful but risky because it’s a black box. You can’t always explain why it behaves a certain way.

That’s why verification and validation (V&V) become critical. You can’t afford hallucinations when an AI driver makes life-or-death decisions in milliseconds.

Where does NVIDIA come into this picture?

NVIDIA is one of the world’s leaders in simulation, sensors, and AI infrastructure. We complement that with safety and verification. The integration creates a full loop: simulate → train → test → verify → deploy. Together, we’re accelerating the industry’s ability to train and validate AI drivers at scale.

NVIDIA was already an investor in our Series C round, but this step takes it further - it’s operational. Our tools now embed NVIDIA’s Omniverse and Cosmos technologies, making it possible to test perception-level AI in realistic, physics-based environments.

Take us back to the beginning. How did Foretellix start?

I started my career at Intel, developing simulation and verification tools for microprocessors. Later, I co-founded Verisity, which completely changed how the chip industry tests and validates semiconductors before production. That company was acquired by Cadence, and the methods we developed are still used by every major chipmaker.

After decades in that world, my co-founder Yoav Hollander and I asked: where else does this approach apply? We looked at robotics, drones, and then at autonomous vehicles. The parallels were obvious. Extreme complexity. Huge cost of failure. And also a massive opportunity to improve safety.

In chips, a bug costs money. In autonomy, a bug can cost lives. And we’re here for that – to help save lives.

I’ve heard you use the term “taming infinity” before. What do you mean by that?

The number of possible driving situations is infinite. You can’t test all of them. So we break the problem into a finite, structured set of scenarios. The ones that matter most for safety. For example, an unprotected left turn with a pedestrian crossing in the rain while a police car approaches. These combinations are endless, but we generate millions of variations that represent that infinity.

You can’t test infinity - so you have to tame it. We use real-world recordings and then create synthetic variations to cover all the edge cases. That’s how we make sure the AI learns from both real and simulated experiences.

How has the industry evolved since you started the company in 2018?

Dramatically. When we started, everyone was focused on building demos - showing that a car can drive down a street. Now, the conversation is all about safety at scale. The last ten percent of autonomy, the edge cases, takes ninety percent of the effort.

As the industry matured and moved from level 2 systems to higher levels like level 3 and level 4 (robotaxis), the need for deep verification skyrocketed. At level 2, the human driver is still liable. At level 3 and above, when it is a hands-off situation, the OEM becomes responsible. That changes everything.

And with the rise of AI - especially end-to-end AI - we had to expand. Verification alone wasn’t enough. So we added tools for training and evaluating AI systems, which is what led to this new partnership with NVIDIA.

What types of customers are you working with today?

Our customers range from robotaxi developers to OEMs and Tier 1 suppliers — in the U.S., Europe, Japan, Korea, and China. We also support autonomous trucking companies that operate on highways. Anywhere AI drives physical motion, we can help.

Right now, we’re also seeing growing interest from robotics and industrial automation. Physical AI isn’t just about vehicles — it’s about machines that interact safely with the real world.

What do you see happening by 2030 or 2035? Will we all be riding in autonomous vehicles?

Not yet — but we’ll get there. By 2030, most new vehicles will include advanced autonomy features. Robotaxis and autonomous trucks will operate in defined ODDs - urban cores, ports, highways. Human drivers will still exist, but increasingly as supervisors. I think full Level 5, where there’s no steering wheel, is still a decade or two away.

But the bigger shift will be cultural: trust. Trust in AI systems that move through the same physical space we do. And that trust must be earned through measurable safety.

And what’s next for Foretellix?

We’ll continue leading the way in safety for Physical AI. The NVIDIA partnership is a big milestone in the middle of our journey. We’re expanding globally, signing new strategic customers, and preparing to enter new verticals like robotics and industrial automation.

Physical AI will power everything that moves — and our job is to make sure it’s safe.

Lastly, before we end - what advice would you give to other founders building deep-tech companies in this space?

Be patient, be brave, and take bold decisions – and be ready to fail. You’ll hear ninety-nine “no” before one yes. Remember - you just need that one yes. That’s how startups survive. Keep a solid ego – to stand up to every “no” and keep going!

#movingpeople is a weekly mobility & delivery newsletter - from Ride-Hailing & Robotaxi through Remote-Driving & Delivery Drones to e-bikes and Autonomous Freight & Logistics.

Thank you for reading. If you like what you're reading, please share it with your friends and colleagues so they can benefit from it too.