Gopal Sharma

Played a significant role in enhancing high-precision Driver Monitoring Systems (DMS) for automobiles and fleets by developing functionalities like image classification, object detection, segmentation, video analysis, face landmark, and face recognition. Achieved 95% accuracy on individual AI models, contributing to an impressive 90% overall accuracy. Detected various driver behaviors including drowsiness, distraction, seat belt usage, phone usage, yawning, rash driving. Collaborated across departments, leveraging edge computing, and optimized CPU utilization. Implemented image, video compression & enhancement techniques, employed parallel computing and threading to minimize latency and inference time, resulting in a remarkable 50% improvement in computational efficiency.

Worked on Deep learning algorithms and CNN Architecture used to solve business problem. Set Up on Optimisation techniques, build model which are compatible with edge devices. Constructed Energy Load Forecasting on time series data of ASX energy shares. Designed and implemented an end-to-end detection & digit recognition model pipeline which extracts consumption of Energy meter from its image with working accuracy of 97% on field test. Optimised the size of pipeline from 1.2GB to 50MB, while maintaining 93% accuracy so that whole pipeline implemented in android application. Developed an impactful payment prediction system that accurately anticipated late or missed payments with a commendable 92% accuracy for the upcoming month. Gathered, cleaned, and processed extensive datasets using MLflow and Airflow, constructing and refining the prediction system. Streamlined processes, resulting in a notable 10% improvement in monthly collections, and facilitated seamless tracking, management, and orchestration of ML workflows.

Implemented an electricity theft detection system using consumer consumption patterns with a 90% accuracy rate. Used rule based and unsupervised learning technique to trained the system, helps to reduce the electricity loss without actually going on field. Communicated and collaborated with field people to get inside of this, build a interactive dash board for EDA to visualised and understand data.