Summary
Punit Kumar is an AI research intern and data scientist who blends rigorous academic training with hands-on ML deployment across logistics, fintech, and consumer platforms. He is pursuing an MS in Artificial Intelligence at the University at Buffalo, after earning a dual B.Tech/M.Tech from IIT Kharagpur. In industry, he has designed and deployed predictive models and recommender systems on AWS SageMaker, including an LSTM/Attention availability predictor for BlackBuck that achieved 84% accuracy and 81% recall, boosting bookings by about 4%. He also built a Transformer-based credit-default classifier (89% accuracy, 81% recall) and led chat fraud detection with a BERT-based multi-label classifier delivering 95% accuracy at Quikr, along with loan-prediction and personalized recommender pipelines. Currently at Decompute as AI Research Intern (since May 2025), he continues to translate research into scalable production workflows. Outside work, he has contributed to IIT Kharagpur's alumni initiatives as an advisor and web lead, reflecting leadership and a broad GitHub footprint across BlackBuck, Voicezen, and Quikr.
10 years of coding experience
7 years of employment as a software developer
Dual Degree(B.Tech & M.Tech), Dual Degree(B.Tech & M.Tech) at Indian Institute of Technology, Kharagpur
English, Hindi, Sanskrit