Now an associate professor in Agricultural and Biological Engineering, Chaterji is the lead author of a patent-pending system called ApproxDet, a content and contention-aware approach to approximate object detection. Designed for real-time performance, the system adapts based on what’s in the camera frame and how much computational power is available at that moment.

The result? A more efficient, flexible method for video analytics that cuts latency in half and improves accuracy on mobile hardware. No easy feat in a world increasingly reliant on edge computing.

Chaterji’s system stands out for its multi-branch detection model which allows it to scale performance dynamically based on scene complexity. An integrated tracker helps reduce the need for full object detection on every frame while conserving device resources without sacrificing quality. The framework also factors in real-time CPU and GPU load while maintaining performance even when mobile hardware is under stress.

While ApproxDet’s immediate applications point to video and smart city surveillance, Chaterji’s vision goes far beyond that. Through her work with the Wabash Heartland Innovation Network she’s helping bring Internet of Things (IoT) tools to the field of agriculture and aiming to make farming more high-throughput data-driven and sustainable. On another front she continues to explore predictive models in genomics and epigenomics for personalized healthcare.

A proud Boilermaker, Chaterji earned her PhD in Biomedical Engineering at Purdue where she received the Chorafas International Award and the College of Engineering Best Dissertation Award. She completed postdoctoral training in biomedical engineering at the University of Texas at Austin with support from the American Heart Association. Today she continues to bridge research and commercialization as a scientific advisor to the IC2 Institute at UT Austin and a recipient of Purdue’s Seed for Success Award.

From agriculture to AI, Chaterji's work shows what’s possible when engineers lead with purpose and let data do the talking.