They include manufacturing automation specialist LaunchPad Build AI and Amazon for its Ring home security cameras, and said it had established partnerships with Nvidia and neo cloud firm Nebius.
So what does Antioch’s software do?
AI-powered robots, manufacturing and any kind of “physical AI” have become huge. Mellsop founded Antioch because the process of testing new models was “absurdly manual” and bursting at the seams.
“We’ve talked to teams literally renting Airbnbs just so they can test their household robots overnight, or spending millions building fake warehouses and neighbourhoods to simulate the real world. It’s slow, expensive and wildly inefficient,” he told the Herald.
Replacing actors
“Even the likes of Amazon Ring have relied heavily on actors and real-world scenarios to train and test their devices,” says Mason Beakley, principal at Icehouse Ventures, which participated in the raise.
Like its rivals, Ring has been adding features such as support for natural language queries, so you can quiz your cam about what time a courier arrived, or your teens came home.
“Rather than physically running those scenarios over and over, Antioch can build a virtual porch and a virtual person, run 10,000 scenarios in seconds, identify failures, change the software, and rerun the simulation,” Beakley said.
“Antioch is already proving that AI simulation can work in the real world, and if it works for something like Ring, the possibilities of what comes next are endless.”
Most firms have been using software simulations alongside real-world testing, but it’s often clunky, Mellsop says. Antioch’s platform boosts the virtual testing.
Nvidia, Palantir execs back raise
The US$32m ($54.6m) Series A was led by top-tier Silicon Valley VC Greylock Partners (an early investor in Facebook, Airbnb, LinkedIn, Figma and Palo Alto Networks, among others) and supported by early existing investors including A*, Category Ventures, Box Group and Icehouse Ventures.
Three tech industry figures also chipped in to Antioch’s Series A raise: Nvidia executive Ian Andrews, Palantir chief technology officer Shyam Sankar and Foxglove founder Adrian Macneil – another Kiwi ex-pat making it big in AI. Macneil – a one-time software developer for Treasury in Wellington – recently raised US$40m ($68.07m) for his startup, which also involves physical AI.
Antioch’s pitch is that it offers the same kind of virtual testing tools used by behemoths like Tesla and Waymo – but at a price affordable for start-ups.
“At the moment we have the ‘good’ problem that we have more opportunities than we can responsibly take on,” Mellsop said.
Antioch has 12 staff and is looking for 15 more to take its headcount to 27.
Mellsop previously said that AI allows a start-up to lean. He says that’s still broadly the case. But when it comes to signing corporate customers; “They want to see people on the other end of the table”.
Some capital will also go to opening a West Coast office, product development and AI compute costs, which are heading north.
The clients onboard so far make for quite the brag list.
“Antioch is changing how we develop and validate complex systems by enabling us to rely on simulation with a level of confidence that wasn’t previously possible”, said Jason Mitura, Amazon’s vice-president for software development and the tech giant’s chief product officer for its Ring security cams.
Part of the Antioch team. From left: Jackson Vachal, Josiah Jordan, Michael Calvey, Colton Swingle, Alex Langshur, Harry Mellsop, Collin Schlager and Oliver Dyer.
“Antioch’s simulations have closely matched our physical test results, including in scenarios we deliberately held out of calibration.
“That confidence lets us move more testing and development into simulation, reducing reliance on costly physical test programs, accelerating engineering cycles, and allowing our teams to focus on delivering better products for customers”.
How did Icehouse get a foot in the door for the Series A?
Mellsop worked for the Auckland VC firm between high school (St Kentigern) and leaving New Zealand for a scholarship at Stanford – which in turn led to a role as a vision engineer in California for Tesla’s physical AI team in 2022.
He then moved to New York, where he co-founded a blockchain company, Transpose, with fellow Stanford grad Alex Langshur.
In May 2023, the pair sold Transpose – still only months old – to the market leader in its field, national security contractor Chainalysis, in a deal the Herald understands ran to eight figures – and co-founded Antioch.
The founding Antioch team also included Colton Swingle, previously at Google’s DeepMind AI unit, and Collin Schlager, previously at Meta’s Reality Labs.
Helping with onshoring, supply chain security
The trio saw an AI business opportunity in founding Antioch, but also a political purpose. Langshur earlier told the Herald that increased AI automation in manufacturing will help re-onshore manufacturing. This week, Mellsop added that it will also boost supply chain security in an increasingly unstable world.
“It’s increasingly clear that Physical AI and robotics are hitting a key inflection moment,” said Greylock general partner Saam Motamedi.
“Software-speed innovation in the physical world will be the foundation of the next industrial revolution.
“We believe Antioch is building the core development platform for this next phase of physical AI.”
Motamedi has also joined Antioch’s board.
“Saam is a really phenomenal VC operator. He also understands us on the technical level,” Mellsop said.
“Finding the right partner for a Series A is one of the most important decisions for the lifecycle of a start-up because you could be working with them for the next 10 years.”
Human networking
Old-school human networking has already played a big role for Antioch, even if its founders have measured each phase of their super-accelerated careers in months rather than years.
Take the investment from Nvidia’s Andrews.
Mellsop and Langshur first met Andrews when he was chief marketing officer for Chainalysis as it bought their start-up Transpose. Andrews went on to become chief revenue for Groq, a LPU (language processing unit) AI chip start-up bought by Nvidia for US$20 billion ($34.03b) in December last year.
Robots, robots, robots
Meanwhile, AI robotics seems to be reaching a tipping point. Every day, social media feeds are filled with humanoid robots working in car factories or, at the recent World Humanoid Robot Games – where bots finished the 100m dash at above world-record pace, even if they comically cannoned off a crashpad after crossing the finish line.
“That wouldn’t have happened if they used Antioch,” Mellsop quipped.
Chris Keall is an Auckland-based member of the Herald’s business team. He joined the Herald in 2018 and is the technology editor and a senior business writer.