Jul 2023 – Mar 2025 · 1 yr 8 mos · Gurugram, Haryana, India · Hybrid

Jul 2021 – Jul 2023 · 2 yrs · Gurugram, Haryana, India

• . Worked on multiple AI Image processing Algorithms to create Virtual Studio look of car listings on OLX Autos.
• Developed car background removal model using salient object detection based on U2Net.
• Developed car parts and damages detection Model in Python using FasterRCNN.
• Built propensity model of churned car seller leads to identify high potential customers. This project improved conversion of call center outbound campaigns by 5%.
• Worked on POC to build car engine audio noise classifier.

Mar 2020 – Jun 2021 · 1 yr 3 mos · Gurugram, Haryana, India

• Fareportal operates several top Travel E-Commerce Websites CheapOAir, OneTravel in North America for Air, Hotel, and Car Products.
• ● Developed Hotel Recommendation Algorithm based on Laplace Score to recommend top Hotels for Cross-sell flows. Improved Cross-Sell Booking Conversion Rate by ~20%. Deployed as a Batch Job using Pandas, Apache Airflow, and MongoDB.
• ● Improved Flight Ranking Algorithm to rank order the flights on Listing Page. Improved Search-to-Book Conversion Rate by ~5%. Used PySpark for Data Transformation and CatBoost model to predict best flights. Deployed on Python Backend using Docker, AWS Elastic Container Service,
• Lambda, and DynamoDB.
• ● Created an Analytical model to identify best flight deals from Internal Search Results/Third-party data. The Project generated $1 million+ in revenue by powering the Price Comparison widget on UI, personalized Deals, and other marketing purposes. Created Python Backend with Pandas, MySQL, Airflow, and MongoDB to integrate it with Marketing Platform. Also created Rest API using Serverless in Python, deployed on AWS Lambda.

Jan 2020 – Jun 2021 · 1 yr 5 mos

• Responsible for Teaching, Doubt resolution, and mentoring the students of PG Diploma in Data Science.

Jan 2019 – Mar 2020 · 1 yr 2 mos · Mumbai, Maharashtra, India

• Athena’s Owl is an AI SaaS for the media domain. It helps media organizations analyze Content and generates Metadata for Business Insights, Cataloging, and Advertisements.
• ● Entity and Brand Logo detection: Implemented Custom Object Detector using Transfer Learning based on RetinaNet to detect various Soccer entities and Sponsor Logos. Used for measuring Brand visibility for Bundesliga (German Soccer League). Efficiently handled class imbalance using the Focal loss function to achieve 95% precision.
• ● Camera Shot Angle recognition: Implemented Camera Shot Angle Identification on Soccer Match Videos using ResNet for Bundesliga. Analyzed False-positive patterns and improved Robustness using Negative classes to achieve 98% Accuracy.
• ● Apparel detection: Implemented Object Detector for Identification of common Apparel and accessories classes in Media Content for a Major US Media Organization.Created a hierarchical Taxonomy to Balance Precision and Class level granularity.
• ● Face Recognition: Implemented Unsupervised Data Clustering for Faces using Graph Convolutional Clustering.Decreased Clustering time and Computational cost to 8% of the existing process.
• ● Built Docker Containers, Established MLFlow Platform to manage Experiments, Models, and Metrics.

Jan 2018 – Jul 2018 · 6 mos · Bengaluru Area, India

• ● Analyzed Geospatial satellite imagery datasets to develop predictive models for cropping/vegetation patterns, water moisture, soil quality using Spectral Indices Analysis.
• ● Improved Land Cover Crop Classification on Landsat 8 Satellite Data using Tree Ensemble Classifier (XGBoost). Achieved 90% precision.
• ● Worked on Agricultural Plot Boundaries Detection on Google Maps using HED Edge detection and morphological Image processing with appreciable results.