This blog post was originally published at Qualcomm’s website. It is reprinted here with the permission of Qualcomm
Enterprise security is not just big business, it’s about keeping you safe: Here’s how engineer-turned-CTO Sudhakaran Ram collaborated with us to do just that.
Key Takeaways:
-
On-device AI enables superior enterprise-grade security.
-
Distributed computing cost-efficiently enables actionable real-time insights.
-
Surveillance cameras computing AI on the edge are bringing differentiated, scalable, effective solutions to market.
The office, a school, the mall, the post office: These are just a few of the places you might visit on any given day. But in an era where safety concerns loom, the need to protect public spaces has never been more critical.
Addressing this is Sudhakaran Ram, CTO and co-founder of FocusAI. His company is paving the way for a new realm of artificial intelligence (AI) security experiences. Leveraging on-device AI, the company is elevating safety with a compelling new solution to a pressing issue.
But AI-based enterprise security is an ever-evolving, competitive market: So how is this brand-new tech venture winning? Ram talks his inspiration, his approach and his most important learnings during FocusAI’s first go-to-market product development cycle.
Meet the startup changing the video intelligence game
Founded in August 2022 in the Bay Area, FocusAI offers sophisticated AI to the landscape of enterprise-grade security. As Ram shares, “FocusAI’s solutions offer real-time actionable insights, privacy protection, and user-friendly interfaces, ensuring a seamless experience for end-users.”
With cutting-edge on-device AI and machine learning (ML) algorithms, the venture-capitalist-backed company offers features not previously possible with cloud-dedicated systems.
Advanced capabilities include gun, face, person and vehicle detection, license plate analytics and more. The FocusAI Dome Camera, along with the Qualcomm Edge AI Box Solutions by Qualcomm Technologies, Inc. is designed to unlock AI’s full potential so an enterprise can keep individuals and assets safe and secure.
The vision: Fast, actionable insights
Ram and his team developed this vision when they noticed a security need for real-time processing and actionable insights.
Knowing that existing solutions back then could take up to 45 minutes to generate AI output, they wanted to provide an easy-to-use on-device AI platform that could generate instantaneous analysis. However, this dream had to address the technical realities. Ram explained, “The cost to compute on the cloud has tremendously increased. Raw video data has high clarity and fidelity, compared to encoded data sent to the cloud that loses quality: This is why processing rich video data needs to happen at the edge.”
Knowing that, FocusAI designed its cameras around a distributed computing model, where the heavy-lifting 7B-parameter large language model (LLM) runs on the built-in QCS8550 in the Qualcomm Edge AI Box rather than the cloud.
Their forward-thinking solution yielded an enterprise security field of dreams: real-time processing, valuable insights and a user-friendly experience… in a more cost-effective wrapper, especially when compared to traditional, cloud-dedicated models.
A mini-case study: Enterprise-grade on-device AI
Recently, a large corporate complex implemented the solution to enhance its traditional system. Ram explained, “They had a traditional system that was mind-numbing: It just recorded and if an incident happened, the staff would need to manually review the video footage.”
The company deployed FocusAI cameras with advanced capabilities to track their complexes in real-time and gain insights into the flow of people throughout the day.
Within a few weeks, the complex was able to identify and address several issues:
- Theft, including costly, ongoing catalytic converter cases
- Unauthorized access
- Misuse of the facility
- Security gaps and other issues
Ram reports their client is already planning to expand the implementation; he added, “We delivered insights in real time so they could review what happened instantly. And they tell us this is very powerful.”
This refined level of processing and camera footage may arguably mean millions of dollars in a shopping mall slip-and-fall claim. While a traditional system will often capture the fall, it will more commonly miss the critical moment: the cause.
FocusAI solves this problem with edge AI processing, which slows down footage up to eight frames per second, allowing you to see moments in sequence so that you can truly determine what happened.
A daunting, yet real-world example would be the threat of an armed person on location. Traditional systems often miss weapons or can’t track them accurately due to lower processing and resolution. With on-device AI, FocusAI can decrease the possibility that a weapon, for example, is camouflaged by a same-colored jacket, so first responders can act faster.
On-device AI spec vetting considerations
As an experienced system-on-a-chip (SOC) designer and former member of the Google Nest team, Ram approached design with a deep understanding of system knowledge and the importance of edge compute capabilities.
He had worked with Qualcomm Technologies’ products and was aware of our capabilities. Nonetheless, Ram conducted a thorough partner evaluation selection process.
While Qualcomm Technologies and competitors had similar compute levels, Ram concluded that our powerful central processing unit (CPU) subsystem, coupled with our graphics processing unit (GPU), positioned us  as the superior partner choice for scalability, features and applications.
The result: a groundbreaking AI security solution that surpasses what others have achieved, setting FocusAI apart.
Learnings from FocusAI
- Real-time monitoring and analysis can address end-user issues.
- On-device AI provides actionable insights at a large scale.
- A comprehensive partner evaluation process can result in superior go-to-market differentiation.
- Performance, system solution capabilities, scalability and tool chain quality are considerations that can align a partner with the unique needs of the device.
- Emphasize scalable ease of use across the design: A solution will not be adapted unless it’s user-friendly.
Lara Piu
Editor in Chief of OnQ and Marketing Manager, Qualcomm Technologies, Inc.
