Vedant JHUNJHUNWALA

Developed a real-time office occupancy detection system to determine chair usage status (occupied/vacant). • Engineered event-based tracking to log precise timestamps for occupancy initiation and release.

Building a system using Variational Autoencoders (VAEs) to detect adversarial attacks on image classifiers. • Exploring VAE and GAN-based defense mechanisms to enhance AI model robustness and security, with scalable

Partnered with clients to deliver AI-powered solutions; • Worked on LLM alignment, prompt evaluation, and model optimization; refined LLMs by designing/testing prompts, enhancing coherence, quality, and safety; • Achieved measurable improvements in client deployment.

Real-time traffic violation detection (~100% accuracy), piloted in Roorkee, scalable pan-India • Built predictive trajectory models for vehicle movement forecasting (Scheduled to be publish in IEEE Journal)

Designed and delivered a real-world data analytics project on IPL datasets using Microsoft Power BI, uncovering key performance trends, team metrics, and player insights. • Created interactive dashboards and visualizations to support data-driven decision-making in sports analytics.

Group Leader - Led a global cohort project, applying data analytics, decision models, and data mining. • Built predictive regression models, clustering (K-means, hierarchical), and association rules (Apriori, FPGrowth).