Srivenkata Gantikota

User performance analysis and correction for S/W

Abstract:

Optimizing user interaction with complex software is a critical yet often overlooked challenge in both healthcare and technology domains. This talk presents an AI-powered framework for analyzing and improving user interaction patterns within software applications, with a primary focus on medical imaging systems such as PACS (Picture Archiving and Communication System) viewers.
The proposed system collects time-series interaction data from users across varying roles and task types. Using supervised deep learning models — including Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), and Time-based Convolution Networks (TCN) — the system learns optimal interaction sequences associated with expert-level users. Once trained, the model evaluates incoming interactions in real time, assigns performance ratings across speed, accuracy, and efficiency, and generates personalized recommendations to improve suboptimal workflows.

A key contribution is the automated establishment of ground truth through software requirement mapping and sequential activity logs, significantly reducing manual labeling effort. The system also supports continuous retraining using outcome-based feedback, ensuring the model adapts as workflows evolve.

Validation in radiology workflows demonstrates measurable improvements in diagnostic efficiency and accuracy. This framework has broad applicability across intelligent systems, IoT, and smart computing environments where human-software interaction quality directly impacts performance outcomes.

Profile:

I am a Software Developer at the University of California, San Diego (UCSD), where they oversee mission-critical financial aid applications bridging rigorous engineering with real-world institutional impact. A prolific IEEE author and Fellow of THREWS, they have served as a reviewer and judge for numerous conference submissions. I also hold a U.S. patent on User Performance Analysis and Correction for Software Systems — a framework enabling software to intelligently detect and respond to user behavior patterns to improve efficiency and outcomes.