As an embedded systems authority, Abhay Mangalore has almost 20 years of experience. He’s at the helm of a revolution to change security cameras forever by introducing cutting-edge artificial intelligence (AI) technology. His novel approach enables quicker security response times. Simultaneously, it takes on key issues such as bandwidth requirements and privacy threats. Mangalore’s work takes advantage of low-power hardware acceleration to increase efficiency on AI tasks as in the example of object classification. He terms this adaptive AI scaling, and it’s what powers these processes to run only when they’re needed.
Mangalore’s journey has not been without challenges. His teams have overcome high mountains developing, optimizing and perfecting AI models for wide-scale deployment at night, in harsh weather conditions, and through other challenging environments. It was his tireless commitment that resulted in the development of numerous award-winning smart home devices. Consequently, Arlo’s camera lineup has won six successive CES Innovation Awards and three successive Red Dot Design Awards. Mangalore imagines a future where the future of smart security is on-device AI. This breakthrough holds potential to improve user experience while safeguarding data as never before.
The Power of On-Device AI
Mangalore’s operating conviction is that inverting intelligence to the edge removes latency headaches and reduces bandwidth consumption. This method offers the ability for data to be processed locally, providing a tremendous boost to the privacy of user data. Security cameras should make people feel safer—not worried about where their data is going. That’s why privacy-first AI solutions are at the very top of our priorities,” he continued.
His strategy includes encrypting any communications, no matter what type, validating any firmware updates and limiting the amount of data shared externally. Mangalore thinks privacy is important. More importantly, he creates it as a defining feature of both the design and experience of his inventions. As he writes, “Privacy is not an afterthought—privacy is built-in from the ground up. Here’s how we ensure data protection: on-device AI processing, where most video analysis happens locally to reduce cloud exposure, and end-to-end encryption, where all data is encrypted using AES-256 and TLS protocols to prevent unauthorized access.”
Additionally, Mangalore’s work is focused on significantly decreasing false alerts and giving users more actionable alerts that are relevant to them. By integrating advanced AI capabilities into compact hardware, he believes many traditional shortcomings of security cameras can be addressed effectively.
Overcoming Challenges in AI Integration
Over the course of his career, Mangalore has addressed many of the obstacles related to AI implementation in various landscapes. “One of the biggest challenges we faced was getting AI models to work reliably across all kinds of environments—from sunny outdoor spaces to dimly lit rooms,” he remarked. His team doesn’t stop there, instead they constantly optimize each individual piece for efficiency and accuracy.
Adaptive AI scaling is a central part of Mangalore’s approach. He explains, “It’s all about smart resource management. Running AI 24/7 on a battery-powered camera would drain it in no time, so we use strategies to make AI as efficient as possible.” For example, Mangalore employs techniques such as model quantization to improve performance. The process quantizes AI models, converting them to lower precision formats that allow them to run much faster without losing accuracy.
“The secret sauce, he admits, is in identifying the balance between performance and power consumption. The goal is simple: maximize performance while keeping power consumption low because security cameras need to be reliable without draining the battery too quickly,” he emphasized.
The Future of Smart Security
Looking forward, her hope is that these best practices will come together across industries, ushering in a new generation of on-device AI. He’s keenly aware of emerging opportunities afforded by faster connectivity technologies, such as 5G, which can facilitate more responsive, real-time security systems. Beyond the hype, he brings home the fascinating promise of multi-modal AI. This innovative technology enables cameras to connect vision, audio and embedded sensor data, unlocking deeper security capabilities.
“The next few years will bring some exciting breakthroughs in security AI, including 5G-powered AI that will enable faster connectivity and more responsive security, and multi-modal AI,” he stated. Both Walsh and Mangalore are optimistic that the industry is changing. It’s moving from simply detecting threats to actively preventing them before they even happen. “Security isn’t just about detecting threats anymore. It’s about preventing them before they happen. That’s where the industry is headed, and I’m excited to be part of it.”
