For millions of people living with diabetes, checking blood sugar is a daily struggle. Some rely on finger pricks — small but painful reminders of their condition. Others use continuous glucose monitors (CGMs), which require skin patches, tiny needles, and expensive sensors that need replacing every few days. For years, the idea of a painless, needle-free way to monitor blood sugar felt impossible.

That is, until Abhinav Agarwal, a Stanford-trained AI expert, helped change the game.

As the Head of Machine Learning at KOS AI, Abhinav played a key role in developing Argus, a first-of-its-kind non-invasive glucose monitoring wearable. Instead of relying on needles, Argus uses light and AI-powered analysis to measure glucose levels through the skin. No pricking, no patches, no discomfort. Just a sleek device worn on the wrist, working in the background to track blood sugar in real time.

The Journey Behind Innovation

When asked about his breakthrough moment, Abhinav explains that it wasn’t a single flash of insight but a series of small victories.

“We were analyzing light data from our early prototypes,” he says. “The glucose information was buried under much stronger signals from the body. Traditional methods couldn’t extract it reliably.”

The turning point came when his team developed a new approach. “When we created algorithms that could analyze relationships between different light wavelengths, hidden patterns emerged. That’s when I realized we weren’t just improving existing approaches — we had found a completely different way to extract glucose information from light signals.”

Abhinav describes his solution with a simple analogy: “Imagine trying to hear a specific conversation in a noisy restaurant. Instead of just trying to filter out background noise, we taught our system to recognize patterns in how the conversation sounds from different listening points around the room. While no single point gives you a clear signal, the pattern across all of them creates a unique ‘fingerprint’ of what you’re looking for.”

Impressive Results Through Innovation

The KOS Argus has already shown promising results in testing:

  • 91.3% accuracy in identifying glucose ranges
  • 93.5% sensitivity for detecting low blood sugar
  • 97.2% specificity for low blood sugar detection

These numbers represent real hope for millions. With diabetes affecting over 537 million adults worldwide, Argus could transform daily life for many who currently endure 1,000 to 1,500 finger pricks annually.

“The KOS Argus transforms glucose monitoring from a painful, intermittent process into a seamless, continuous experience,” says Shalal Habib, CEO of KOS AI. “Our technology eliminates the pain and inconvenience of finger pricks while providing reliable glucose readings.”

Education and Background

Abhinav’s path to creating this breakthrough technology was shaped by his education at Stanford University, where he studied Computer Science with a focus on Artificial Intelligence.

“Stanford’s approach encouraged collaboration across different fields,” Abhinav says. “The breakthroughs behind Argus emerged at the meeting point of machine learning, signal processing, optics, and medical science.”

His education provided more than just technical knowledge. “Stanford connected me with an extraordinary network of collaborators and mentors. Several key members of our technical team are fellow Stanford alumni, and we maintain active research collaborations with the university.”

Before joining KOS AI, Abhinav worked as an AI Systems Developer at GDX Co., Ltd. in Tokyo and as a Machine Learning Engineer at organizations including Krungthai Bank. He has also served as an AI Instructor at Khan Lab School and as an AI Research Mentor at Inspirit AI.

Beyond his professional work, Abhinav founded TechSoCh (Technology for Social Change) and co-founded Young Reformers Commune, initiatives that have provided education to over 230 underprivileged children across multiple countries.

Award-Winning Research

Abhinav’s talents extend beyond glucose monitoring. His research on making large AI models more efficient won Best Paper at the ICAIRME Conference 2025 in Bangkok.

“We developed techniques to systematically identify and remove redundant parts of AI models,” he explains. “Our approach achieved a 35% reduction in model size and 40% improvement in speed while maintaining 94% accuracy on healthcare tasks.”

This research directly benefited the Argus device, which needs to run sophisticated analysis on a small wearable with limited battery life.

Overcoming Major Challenges

Creating Argus required solving several difficult problems that had stumped previous attempts at non-invasive glucose monitoring.

“The first major challenge was the extremely low signal-to-noise ratio,” Abhinav explains. “The optical signatures of glucose are incredibly subtle and easily overwhelmed by variations in skin properties, blood flow, temperature, and movement.”

His team also had to ensure the device worked for everyone. “Factors like skin tone, tissue thickness, and hydration levels all affect how light interacts with tissue. Our solution was to develop a self-calibrating system that could adapt to individual characteristics.”

Perhaps most importantly, they had to make it work in real life, not just in a lab. “Traditional approaches work reasonably well in controlled settings but break down in daily life. We implemented a combination of hardware stabilization, motion sensing, and adaptive filtering to maintain signal quality during normal activities.”

Beyond Glucose: A Vision for the Future

While Argus focuses on glucose monitoring, Abhinav sees a future where similar technology could track many aspects of health.

“The core innovations we’ve developed have applications for monitoring many other health conditions and biomarkers,” he says. “What began as an effort to solve one challenging problem has evolved into a broader platform that could transform health monitoring across multiple conditions.”

Abhinav believes we’re entering a transformative period in healthcare. “In the next 5 years, we’ll see significant advancements in continuous health monitoring through non-invasive wearables. These technologies will expand beyond single measurements to track multiple health indicators, providing a more complete view of health.”

He’s particularly excited about making advanced health insights available to more people. “While today’s advanced diagnostics require expensive equipment and specialist interpretation, this technology will increasingly enable powerful health insights through accessible consumer devices. This will be particularly transformative in regions with limited healthcare access.”

A Personal Mission

For Abhinav, this work is more than just a technical challenge. “My own family member lives with diabetes, and seeing the daily burden of finger-stick monitoring has made this mission personal,” he shares. “Knowing our work will free millions from this painful routine and potentially improve health outcomes through better monitoring is profoundly motivating.”

The KOS Argus will be unveiled at a press conference in Palo Alto on March 20, 2025, marking an important milestone in health technology. Thanks to Abhinav Agarwal and KOS’s remarkable innovations, painless, continuous health monitoring is becoming reality — changing lives and opening doors to a healthier future for millions worldwide.

For more information about Argus and KOS AI, reach out.