Skylark Labs, led by founder and chief executive officer Dr. Amarjot Singh, is betting that a brain-inspired “AI brain for machines” built at the edge can reshape how cities, borders, and critical systems stay safe and operational. From its base in the United States, it is rolling out technology that enables devices to learn directly from their environments in real time, without waiting for massive datasets or cloud retraining, at a time when security, mobility, and infrastructure are facing faster, more unpredictable changes than ever.
The company’s central question is consequential but straightforward: what if machines could decide for themselves what to learn, what to remember, and what to forget, much like the human brain does? Skylark Labs’ answer is a self-learning AI brain that runs on-device, coordinates across a network of machines, and adapts to new patterns as they emerge in the field, from unusual traffic behavior to new threat signatures at a remote border crossing.
Dr. Singh mentions, “Our key breakthrough is that the system keeps getting smarter. 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.”
What a Self-Learning AI Brain Really Means
For many readers, “self-learning AI brain” may sound like marketing shorthand for yet another algorithm, but Skylark Labs uses the phrase in a literal, technical sense. Its core technology is a brain-inspired hybrid architecture that mimics how the human brain processes information, combining fixed components, unsupervised learning, and supervised layers so that machines can decide what to learn, when to know it, and what to forget without constant human-driven retraining.
In practical terms, that means the AI can update itself while deployed, learning from individual examples and reorganizing its own “memories” to stay relevant. The design borrows from how the neocortex and hippocampus interact, separating short-term, fluid memories from long-term, consolidated knowledge, then deciding which patterns deserve reinforcement and which can be discarded. The aim is not to copy biology perfectly, but to capture its most useful principles for machines working in the field.
The models run at the edge, on cameras, towers, and embedded devices, rather than depending on a central cloud server to process every frame or signal. When a system encounters something new, such as a previously unseen vehicle maneuver or a different way people interact with infrastructure, it can learn from that instance and recognize similar events later without waiting for engineers to collect that data, label it, and retrain a model in a data center.
Dr. Singh stresses that this is intentional, describing it as “AI that truly learns on the edge without datasets or retraining, an AI that works like a human brain, deciding for itself when to learn, what to learn, and how to integrate that knowledge.”
A crucial part of this self-learning concept is the ability to forget. Traditional AI systems often accumulate parameters and patterns indefinitely, leading to memory bloat, higher compute costs, and degraded performance over time. Skylark Labs’ architecture introduces automated forgetting, where infrequent or irrelevant patterns are gradually phased out while recurring, meaningful ones are strengthened and moved into long-term memory.
Dr. Singh has compared this to how the brain consolidates memories during sleep, explaining that the system continually evaluates what matters most and “releases what doesn’t,” keeping the model lean and focused. This design marks a deliberate departure from the dominant AI paradigm that demands massive datasets and repeated training cycles. Dr. Singh explains, “The brain doesn’t wait for you to collect a million examples before it understands something, and our architecture is built on that same principle of learning efficiently from very little data.”
Why Industries Need an AI Brain for Machines
The case for a brain-inspired self-learning AI is not only about technical elegance; it is about the environments where Skylark Labs deploys its systems. Its technology runs along borders, in cities, across highways, on campuses, and around sensitive facilities, where new threats and behaviors constantly emerge. In these places, static models trained on historical data struggle because they may fail when confronted with novel tactics, unusual weather or lighting conditions, or rare edge cases that never appeared in their training sets.
When that happens, the traditional response is slow and costly. Teams must collect new data, label it, retrain the model, and push updates, all while the system continues to operate with an incomplete understanding of what is happening around it. That lag creates a gap between what the model knows and what the world is doing. Skylark Labs positions its brain-inspired AI to narrow that gap by enabling systems to learn where they are deployed.
By learning on-device, the brain-inspired AI can adapt to local patterns and adjust its thresholds and predictions accordingly. In public safety deployments, that can mean distinguishing more accurately between harmless gatherings and emerging risks, reducing false alarms and alert fatigue for human operators while preserving a high sensitivity to genuine threats. In traffic safety, self-learning enables enforcement systems to recognize new forms of risky driving or emerging patterns of noncompliance before they show up in national statistics.
Another factor is efficiency. Its hybrid architecture aims to achieve significantly better efficiency in data processing, computation, and energy use compared with conventional models, enabling advanced intelligence in low-power, bandwidth-constrained scenarios. That efficiency allows drones, roadside units, and remote towers to run sophisticated perception and decision-making logic without relying on continuous connectivity to a centralized cloud. This constraint has historically limited where AI could operate. Dr. Singh adds, “If an AI system cannot keep up with the world it is deployed in, people stop trusting it. We wanted to build a brain that stays aligned with reality, not just with the data it saw in the lab.” That on-the-ground mindset is where Skylark Labs sees its self-learning AI brain as most relevant today.
How Skylark Labs Turns Research Into Embodied Superintelligence
The idea of a self-learning brain for machines could have remained a research concept, yet Skylark Labs has built its business around deploying it in what it calls embodied superintelligence—real devices that sense, decide, and act in physical space. Its platforms, Kepler and Turing, sit at the center of these deployments, connecting cameras, radar, LiDAR, drones, and other sensors into systems that can collaborate while keeping data on-site.
In public safety deployments, the company’s AI towers use this brain-inspired architecture to detect threats, learn what normal behavior looks like in a given environment, and reduce noise for human operators. Reports from these implementations describe how the system improves over time by strengthening memories of genuine risks and releasing patterns that do not recur, thereby keeping the focus on incidents that matter. Dr. Singh has likened this process to human memory consolidation, explaining that “just as your brain consolidates important memories and discards trivial details, our system continuously evaluates what to keep and what to release.”
Dr. Singh often describes this collaborative dimension as essential rather than optional. “To be truly intelligent, machines must not only learn for themselves but also learn from each other,” he says. In his view, the future Skylark Labs is working toward is one where embedded systems quietly share insights to help societies respond more quickly and with less friction to emerging risks, while staying within clear operational and ethical boundaries.
That future points back to the promise in the headline. The self-learning AI brain for machines that Skylark Labs is building is not a single monolithic system; it is a growing network of embodied, adaptive agents designed to keep real-world operations safer, more efficient, and more accessible.
By focusing on edge learning, memory, and collaboration, Skylark Labs is trying to move AI from static tools to evolving partners in the infrastructure people rely on every day.