Vanshika Jain

Performed data gathering, mining, and processing on large datasets, resulting in the creation of robust data models of 10K lunar images, providing valuable insights for lunar exploration. Introduced vision to the mechanical robot for scene segmentation tasks using deep learning, and achieved a 0.78 IoU score.

Engineered a supervised ML algorithm using ensemble techniques for multi-class narcotics drug classification and achieved 96% accuracy. Employed peak detection and feature engineering methods to analyze tabular data comprising 80K samples from 15 drugs, providing valuable insights into drug characteristics. Collaborated with cross-functional teams to deploy 30 production systems, researched data model optimization, and conducted root cause analysis to reduce false alarms by 67%. Contributed to continuous improvement using Git to ensure accuracy and presented analytical results to stakeholders.

Conducted comprehensive data analysis to develop machine-learning models for real-time object detection, human pose estimation, and action recognition, improving accuracy by 25% in assisted living technology. Achieved 2x speedup on Intel’s NCS2 and 10x on NVIDIA's Jetson while maintaining accuracy using quantization techniques.