From Vision to $21M Reality: Amarjot Singh’s Self-learning AI Bet at Skylark Labs

From Vision to $21M Reality: Amarjot Singh’s Self-learning AI Bet at Skylark Labs

This New York startup just clinched a milestone contract that could redefine traffic enforcement across one of Asia’s fastest-growing economies, and signals a major shift in where AI’s real value lies.

Photo courtesy of Skylark Labs

When Dr. Amarjot Singh founded Skylark Labs in 2021, traffic AI was stuck in the cloud, bulky, expensive, and too complex for real-world deployment. He saw a better way: build brain-inspired artificial intelligence (AI) that runs directly on police cars and cameras, no cloud required.

Now, his bold vision has become a reality. Skylark Labs has landed a $21 million contract to deploy its Kepler™ platform across 6,000 law enforcement systems in a major Asian nation, one of the region’s largest public safety AI deployments. For the founder, Dr. Amarjot Singh, it’s not just a win. It is proof that simple, scalable edge AI is the future.

“Most AI companies are still stuck in complex deployment approaches for physical AI,” Dr. Singh mentions. “The real opportunity isn’t in sophisticated hybrid systems—it’s in making smart technology simple enough to deploy anywhere, where every second counts for things like traffic safety and national security.”

Skylark Labs’ Edge Advantage: Saving Millions While Preserving Privacy

The Asian uses Kepler™, Skylark Labs’ system that works directly on police cars and traffic cameras to instantly spot traffic violations and dangerous driving. While many competitors need constant internet connections to work properly, Skylark Labs’ system does most of its thinking locally, so it keeps working even when internet service is spotty, a big plus in developing countries where internet infrastructure isn’t always reliable.

Additionally, Kepler™ stands out because every patrol car hosts the full AI stack: rapid violation detection, predictive risk analytics, and a continuous-learning loop that retrains models right on the GPU inside the cruiser. Those self-updating models feed directly into ticketing and CAD/RMS systems, turning each vehicle into a high-margin SaaS seat.

“We built our system to be local on the device from the ground up,” Dr. Singh explains. “Our advantage is that the smart technology works right on the device, which cuts millions in ongoing costs per year for cities and protects privacy better. Every police car becomes a mini-computer that handles traffic monitoring locally. After we set it up, cities spend much less on monthly fees than systems that need constant internet connections.”

“The introduction of this system has provided us with a practical means to improve traffic enforcement across the State. It enhances field coverage for our officers, enables prompt detection of violations, and is helping reduce road accidents on high‑traffic corridors.” Nand Lal, A senior government official with The Ministry of Road Transport and Highways (MoRTH). “The solution is cost‑effective, allowing us to achieve better outcomes without significant recurring expenditure or additional infrastructure. Importantly, improved compliance is augmenting the Government’s non‑tax revenue, and these proceeds are being allocated to programmes that further strengthen road safety and public infrastructure in the public interest.”

Beyond cost savings, Skylark Labs also delivers another important benefit: privacy. While the industry is getting better at protecting privacy with anonymous data, the company’s approach means the important detection work happens right on the device.

Dr. Singh adds, “Privacy protection is built into how we designed the system. The critical processing happens right where the data is collected, giving cities more direct control over their traffic enforcement information.” 

This fits well with stricter privacy laws and growing concerns about being watched. According to IDC’s March 2025 report on edge computing, spending on this technology is expected to reach $380 billion by 2028, driven by the need for faster response times and more reliable systems, especially in places with unreliable internet.

From Cambridge Insights To Asian Roads

Before founding Skylark Labs, Dr. Singh began his journey during his PhD in Mathematics and Neuroscience at the University of Cambridge, where he worked on building brain-inspired hybrid AI systems that learn and think like the human brain. 

Instead of using massive datasets and complex networks, he designed a lightweight AI that processes information like brains do: basic sensing first, learning in the middle, and decision-making at the end. This made the AI fast, efficient, and able to run on small devices, becoming the core of Skylark Labs’ technology.

Later, for his PostDoc at Stanford University, Dr. Singh ran into a big problem: when these AI systems were put into real-world devices, they would see new situations they’d never been trained on, causing them to break down or stop working. 

His solution came through his work on a government program called DARPA Lifelong Learning Machines, where he developed AI that doesn’t just work on devices—it keeps learning while it’s working.

“Our key breakthrough is that the system keeps getting smarter,” Dr. Singh shares. “It doesn’t just spot patterns. It learns and improves in real-time based on what it sees, without needing to be retrained or connected to the internet all the time.”

Skylark Labs’ AI runs directly in police cars, spotting violations instantly, even in unusual cases like overloaded bikes or obscured vehicles. When it sees something new, it learns on the spot. If the pattern repeats, it shares that learning across the fleet. If not, it stays local. This enables fast responses and continuous improvement, without extra costs.

“We’re solving two problems at once,” Dr. Singh explains. “First, we’re making roads safer through smarter, more reliable enforcement. Second, we’re helping governments create steady income streams when they struggle to fund basic infrastructure and safety programs.”

Beyond Traffic: The Bigger Mobility Play

But Skylark Labs’ ambitions extend beyond traffic enforcement. The company has strategically positioned itself at the intersection of several massive markets. 

  • In defense, the Indian Navy is deploying Skylark’s Tracer AI software to scan aircraft carrier decks for foreign object debris. This seemingly mundane task costs military aviation billions annually in damaged equipment and delayed missions.
  • In automotive safety, Mercedes-Benz Research & Development India has partnered with Skylark Lab on an Accident Hotspot Detection System that provides drivers with real-time risk insights and alerts them 500 meters ahead of potential danger zones. The global automotive AI market, projected to reach over $25 billion by 2027, represents another multi-billion-dollar opportunity for Skylark’s edge technology.
  • In infrastructure, Indiana’s Department of Transportation is installing Skylark’s solar-powered Scout towers along major highways, creating a real-time traffic intelligence network that could revolutionize how states manage their road systems. 

Each application shares a common thread: they involve dynamic, unpredictable environments where traditional rule-based systems fail but adaptive AI thrives. This is where Skylark’s edge-first architecture provides a decisive advantage and leads to multi-billion-dollar opportunities in different critical and billion-dollar value sectors. 

Market Opportunity and Existing Competition

The smart-transportation prize keeps expanding: a January 2025 MarketsandMarkets report pegs the sector at US $276.65 billion by 2029—up from $129.7 billion this year—and flags Asia-Pacific as the fastest-growing arena. Singh says Skylark’s new 6,000-vehicle rollout is simply “deal No. 1” in a regional pipeline now at the term-sheet stage, positioning the company to ride that wave.

Competition is real but still siloed. Hayden AI has installed 300 mobile enforcement cameras on New York City buses and is contracted for 140 (with an option for 600 more) on the Washington Metro’s fleet. Rekor Systems sells a Mobile LPR-2 kit that bolts onto cruisers and uses on-device AI to scan thousands of plates per hour. Axon’s Fleet 3 dash-cam ships with a built-in ALPR that claims up to eight times the reads of legacy systems. And Motorola’s M500 couples 4-K video with in-car license-plate recognition and real-time alerts. 

All four are edge-processed, yet none offer Kepler’s live self-training across an entire fleet. That edge focus is where the broader market is heading: Deloitte projects enterprise edge-computing spend to grow about 22% a year, far outpacing traditional IT infrastructure

With this, Dr. Singh argues Skylark Labs’ head start matters. He mentions, “We architected for edge from day one. While competitors are untangling cloud code, we’re already learning from thousands of live miles.” If the numbers and the momentum hold, the adaptive edge could be Skylark Labs’ moat, at least until the big guns catch up.

The Road Ahead

As billions of connected devices come online in the next decade, the companies that can deliver reliable, instant intelligence at the edge will write the rules for the next generation of mobility. For now, Skylark Labs is in pole position—at least in Asia’s fast-growing smart city race.

This frontier extends beyond traffic tickets. The company’s approach could reshape how cities worldwide deploy AI for public safety, infrastructure management, and urban planning, all while keeping sensitive data local and costs manageable. 

Dr. Singh acknowledges that Skylark Labs’ next test will be execution. Deploying AI-powered traffic monitoring systems across 6,000 public safety units isn’t just a technical challenge; it’s a balance between public safety, privacy rights, and municipal fiscal needs.

For Skylark Labs, this deployment represents both validation and opportunity. In a region where infrastructure investment is booming and public safety concerns are paramount, their device-only approach offers a compelling value proposition.

“What we’re seeing is just the beginning,” Dr. Singh says. “The future of mobility AI will be built at the edge, not in the cloud. And we intend to lead that transformation.”