From Robotaxi Fleets to Simulation Labs: One Engineer’s Journey

Robotaxis now run paid routes in a growing list of U.S. cities, from Phoenix and San Francisco to Austin. Autonomous trucks haul ore through mapped mine sites. Driverless tractors work fields with no one in the cab. Each of those depends on a loop that rarely gets described: building models, testing them in simulation and on real vehicles, collecting the data, and validating the safety case, in whatever order the problem demands. Gattani has worked at both ends of that loop, first as the person who decided when a robotaxi was ready for public roads, now building the simulation and data systems the wider industry runs on.

He's seen autonomy from two sides. At Cruise, one of the first companies to put fully driverless cars on public roads across several U.S. cities, he owned vehicle integration and platform readiness, the gate every robotaxi cleared before it could carry a passenger. He now leads technical programs at Applied Intuition, a Sunnyvale company building the digital infrastructure to bring intelligence to every moving machine on the planet, from cars and trucks to mining, construction, defense, and agricultural equipment.

From test tracks to driverless

Gattani trained as an automotive engineer, with a bachelor's from Manipal Institute of Technology in India and a master's from the University of Michigan, Ann Arbor. His first job put him on the test track, driving cars, collecting data, and using it to fine-tune the driving dynamics of what were then Tata Motors' next-generation vehicles, now in production. After that he moved to Cummins, the world's largest independent manufacturer of diesel engines, where he built solutions to cut emissions from the massive diesel engines in commercial vehicles and keep them compliant with regulations across markets.

California was a deliberate break from that track. Rather than move up through an established automotive program, he joined Cruise, then attempting something almost no one had pulled off: fully driverless commercial operation on public roads, no safety driver.

He started at the integration layer, where autonomous software meets physical vehicle hardware — one of the most important and least forgiving parts of the autonomy stack. Gattani's job was to find hardware-software integration failures and drive them to resolution. At the time those failures were the main thing keeping the company from scaling its testing and data collection.

His scope then grew to fleet scaling. He owned the platform-readiness process as Cruise went from roughly 100 hand-assembled robotaxis to nearly 1,000 coming off an automated GM factory line. He wrote the readiness criteria every one of those vehicles had to clear and ran the handoff between the autonomy teams and the factory. When Cruise became the first operator to run fully driverless across several U.S. cities, San Francisco, Los Angeles, Phoenix, Houston, and Dallas, each of those launches cleared the readiness gate he owned.

Every city broke something new. Road geometry, traffic, weather, signage, and local driving habits all changed from market to market, and for each launch his team had to prove the platform met its readiness gates under those specific conditions. The more durable work came after each launch: he turned it into a repeatable playbook, the validation steps, the failure modes to check, the criteria to clear, so each new city cost less time and money to bring online than the one before. The playbook cut the time to autonomous operations in a new city from four or five weeks down to two.

That's the view he came away with. He came to see scaling self-driving as more than a software problem. It's equally a manufacturing, safety, and logistics problem, and a matter of operational discipline.

Deciding when a vehicle is ready

The hardest skill from those years, Gattani says, wasn't scaling a fleet. It was calling when a vehicle was ready for the road, and having the standing to say when it wasn't.

The testing sequence is deliberate. Closed-course testing with a safety driver comes first, then the same closed course with the driver removed, then limited public-road testing with a driver, then constrained driverless operation in lower-risk conditions, often overnight on chosen corridors. Broad expansion only comes after all of that clears.

During the early development of the next-generation platform at Cruise, Gattani's job was to identify and triage the issues blocking public-road readiness and drive them to resolution across several engineering teams, software and hardware alike. When early testing surfaced a phantom-braking issue, the decision on whether the milestone slipped ran through him.

"Every phase had quality gates. If we did not hit them, we delayed," he says. "The platform had to be safe before it went on public roads. That was the standard, and enforcing it was my job."

One visible failure can set a whole program back. Autonomous vehicles run in the physical world, where a mistake carries an immediate safety cost, which is what makes that discipline matter.

"The stakes of not delivering a safe product are extremely high," he says. "In my space, moving fast and breaking things usually means the company breaks too."

Picks and shovels

After several years inside robotaxi work, Gattani made a bet on where the next decade of progress would come from. Not from any one operator's fleet, he decided, but from the infrastructure every serious autonomy program needs: simulation, data, and validation.

"Instead of being in a vertically integrated company that goes all the way from development of the stack to deploying robotaxis on the road," he says, "I wanted to work on the picks and shovels of the industry."

He's now the technical program lead at Applied Intuition, a Sunnyvale company founded in 2017 and now valued at $15 billion, building the digital infrastructure for physical AI. Eighteen of the top 20 global automakers use its tools, along with the U.S. military and its allies and programs across trucking, defense, mining, construction, and agriculture. Its platform runs millions of simulations a month and processes petabytes of autonomy data.

He leads large cross-functional programs that span many of the company's engineering teams. He turns market and customer needs into roadmap priorities, defines what success looks like before work starts, and catches the cross-team dependencies that quietly sink large programs. If one team's milestone depends on another team's unfinished platform work, finding that early is his job. He also built the metrics framework the organization now uses to track progress across teams and judge whether shipped features actually improve customer outcomes.

For many of those customers, this infrastructure is the backbone of development, not a convenience. Simulation and data platforms set how fast a team can test new capabilities, diagnose failures, and validate safety at scale.

The safety case

Safety still anchors how Gattani thinks about the field. Human driving kills and injures enormous numbers of people every year, much of it from distraction, fatigue, impairment, and slow reactions. Developed and validated properly, autonomous systems can cut that by swapping human error for machine behavior that's under constant test.

What motivates him is a measurable drop in injuries and deaths.

He also sees autonomy changing how people and goods move. Robotaxis could lower urban car ownership and hand commuting time back to riders. Autonomous trucks could make freight more efficient. Autonomous mining and farming equipment could take people out of dangerous jobs. Defense systems could handle missions that demand reliability and precision. The simulation and data infrastructure his teams ship to automotive customers feeds those industrial and defense systems too.

From simulation to streets

Gattani's career runs through most of the layers of the industry. He started in conventional automotive engineering, personally owned the integration and readiness gates for one of the first fully driverless robotaxi fleets, ran the readiness process that took that fleet from roughly 100 hand-built cars to nearly 1,000 off an automated GM line across five U.S. cities, and now leads the programs behind simulation and data infrastructure that runs millions of simulations a month and serves customers that include most of the world's top automakers.

Safety at scale is the throughline. For Gattani, autonomy's real payoff is measured in mistakes avoided: machines that can eventually drive, haul, and operate with fewer errors than people, validated by someone accountable for proving it, one readiness gate at a time.