Subhadeep Karan
About Subhadeep Karan
Subhadeep Karan is a Senior Machine Learning Engineer at ACV Auctions, with a strong academic background including a Ph.D. in Computer Science from the University at Buffalo. He has contributed to significant advancements in machine learning applications, including developing a pricing model that streamlined operations and improved auction success rates.
Work at ACV Auctions
Subhadeep Karan has been employed at ACV Auctions as a Senior Machine Learning Engineer since 2019. In this role, he focuses on developing and enhancing machine learning models to improve auction processes. His contributions include addressing the cold start problem through effective feature embeddings and enhancing model transparency by creating real-time performance tracking dashboards.
Education and Expertise
Subhadeep Karan holds a Doctor of Philosophy (Ph.D.) in Computer Science from the University at Buffalo, where he studied from 2014 to 2018. He also earned a Master's Degree in Information Technology from the Indian Institute of Information Technology, Allahabad, from 2011 to 2013. Additionally, he completed his Bachelor's Degree in Information Technology at the University of Mumbai from 2006 to 2010.
Background
Before joining ACV Auctions, Subhadeep Karan gained valuable experience in various roles. He worked as a Research Assistant at the University at Buffalo from 2014 to 2019. He also held positions at SAS, serving as a Cognitive Computing Intern in 2017 and a Summer Machine Learning Fellow in 2018. His early academic experience includes a research position at the Indian Institute of Information Technology, Allahabad.
Achievements
Subhadeep Karan has made significant contributions to machine learning applications. He increased the auction success rate by 9% and improved the acceptance of recommended prices by 30% through model calibration using historical auction data. He developed a pricing model for used cars that streamlined the pricing team, reducing the requirement from 10 full-time equivalents to 2, with a mean absolute error of $500. He also led the development of the open-source ML library SABNA, which has been utilized in real-life applications.