Arpit Dwivedi Email & Phone Number

Bengaluru, KA, IN

Bengaluru, Karnataka, India

Part of the Bing Ads TeamWorking on training models for predicting the relevance between query-ad pairs

Working on training of generative dialogue systems using Reinforcement Learning and development of a novel mutual information based evaluation metric under the supervision of Prof. Pawan Goyal, CSE, IIT Kharagpur

Used a pre-trained Wav2Lip model for generating new lip synced videos of an arbitrary identity to match an input audio segment and improved the performance by restricting lip movements during pauses using OpenCV and Pydub

As a part of the Bing Ads team worked on identifying spoof domains in ads using Siamese Convolutional Neural Networks and achieved a validation ROC-AUC score of 98.93%. Used Faiss to search the nearest legitimate domains and identified spoof URLs in the production system.

Worked on early identification of severe bugs in Ubuntu under the supervision of Prof. Animesh Mukherjee, CSE, IIT Kharagpur

Deployed You Only Look Once (YOLO) algorithm to build a tree detector and classified the trees with an accuracy of 88% using convolutional neural networks. Performed image processing using OpenCV and filtered out the false positives post tree detection leading to an increase in precision.

Built a classifier to distinguish between IT and non-IT jobs with an accuracy of 93% using TF-IDF weighted Word2Vec and XGBoost. Extracted data using SQL and analyzed the trends in the skills and experience of users as well as the types of jobs analyzed.

Minor Degree, Computer Science

Integrated Dual Degree (B.Tech + M.Tech), Electronics And Electrical Communications Engineering

Cbse, Science