• Developed a OCR system for extracting vehicle license plate numbers with high precision. Implementedimage dewarping techniques using Python and OpenCV to correct the skewness angle of input images, ensuringaccurate recognition. Utilized the PP-OCR framework as the backbone for text extraction, combining it withcustom algorithms to achieve Character Error Rate of 0.5 percent• Developed a highly accurate license plate masking system, achieving 98% accuracy in license plate detectionusing the trained YOLOv8 large model. Utilized YOLOv8 weights and the Segment Anything model to generatesegmentation annotations, enabling the creation of binary masks specific to license plates. These masks werefurther combined with logo masks, resulting in an innovative approach to mask the license plates with logos.This system ensured privacy and compliance with sensitive data regulations while also significantly reducingcosts, saving 1000 Rs per car in Photoshop expenses.• Developed a recommendation system for the auction team, enabling sales representatives to deliver personalized recommendations to dealers based on their buying and viewing history. The first version of the solution utilized Cosine Similarity to measure the similarity between inventory and the dealer’s viewing and bidding history, resulting in a 52% click-through rate on the provided recommendations. In the second version, a binary classification approach was implemented, resulting in a remarkable 77% click-through rate on the recommendations. Additionally, 83% of clicked recommendations received bids, and 98% of clicked and bidded recommendations were successfully delivered

*Prepare, process, normalize various whole genome data for the purpose of evaluating the quality of imputation on data.*Experimented with various ways of improving the performance of data access using a non-cluster based environment from delta lake.*Created a Dashboard for the analysis of genetics data-based recommendation(prediction) for groups of traits of disease.

1)Build semantic and syntactical searching system.2) Maintain embedding and data pipeline from data cleaning, geocoding, to statistic feature generation.3) Build a system for recommending the most influential economic indicators using a combination of one vs rest modelling, anomaly detection and curating feature to recommend based on the weighted feature importance of each model.

Exploring and experimenting with various ways for table detection, cell detection, and information extraction from invoice documents for OCR performance improvement.

1)Used advanced augmentation techniques to improve the accuracy of Yolov3 for PPE detection from 81.2% to 91.4%.2) Building intrusion detection system using OpenCV and Yolov3 with current mean average precision(mAP@0.5) of 0.55.

1) Machine Learning and Deep Learning Instructor.2) Recording Interactive Machine Learning, Deep Learning, Mathematics For Machine Learning and Computer Vision Course.

1) Made use of Quandle, Alpha Vantage, and various open economic platform for real-time continuous stock data extraction.2) Performed EDA, Clustering, and Feature Engineering to generate better technical and non-technical features for better recommendations to invest.3) Maintained Dashboard using Grafana for optimum visualization for customers.

● Identifying driving styles using classification models with in-vehicle, data can provide automated feedback to drivers on their driving behavior, particularly if they are driving safely.● Using Machine Learning predictive algorithms and continuous ETL from OBD of vehicle predicting the fuel usage, condition of the vehicle, etc.

● Extracted real-world visual traffic data, performed image segmentation, and then object detection for solving the Loop Closure problem associated and encountered with autonomous vehicles.