About

Subir Verma is a highly skilled Principal Data Scientist with expertise in personalization and recommendation, document information extraction, semantic search and ranking, Q&A chatbots, and image understanding. With over 6 years of relevant experience, he has built and deployed models using Python, Flask/FastAPI, Airflow, and Docker on AWS and GCP. He is also proficient in dashboarding with Snowflake, Streamlit, and Dash. Currently working at Tokopedia, Subir is driving innovation in home, recommendations, and personalizations. He is responsible for various projects, including customer understanding engine, user interest ranking, product de-duplication and diversification, fairness and exposure of products/seller in the marketplace, and similar product recommendation. He also works on vector representation. Prior to Tokopedia, Subir worked as a Data Scientist III at 1mg, where he was responsible for dynamic shipping charge optimization and designing and developing data-schema and metrics pipeline. He also worked as a Data Scientist (NLP) at Shaw Academy, where he developed a chatbot named Morpheus, powered by state-of-the-art NLP models like T5 and BERT. He also worked on enhancing user search with semantic search on product and blogs search page using FAISS indexing, sentence-transformers, T5 model, and FastAPI. He also developed an interactive analytics dashboard for webinar chat and mined important hidden trends and topics from student-agent conversations. Subir holds a Bachelor of Technology (BTech) degree in Digital Signal Processing (Specifically Audio Signal Processing), Circuit Designing & Engineering M from lnmiit and completed his high school from Delhi Public School, Bokaro Steel City. He also attended Navyug Vidyalaya. Subir's tech stack includes Data Scientist, Research Scientist, Python, Search, AWS, Snowflake, Apache Airflow, Docker, Flask, GCP, NLP, BERT, MySQL, and Elastic Search. He also has a Medium blog and GitHub profile where he shares his work.