Vignesh Venkataraman
About Vignesh Venkataraman
Vignesh Venkataraman is a Machine Learning Engineer 2 currently working at Zepto in Mumbai, India. He has a background in data engineering and machine learning, with experience at companies like Willings, Inc., Vahan, and Google Summer of Code.
Work at Zepto
Vignesh Venkataraman has been employed at Zepto as a Machine Learning Engineer 2 since 2022. In this role, he has contributed to various projects aimed at enhancing operational efficiency and improving user experience. He developed a Cross encoder framework using PyTorch and BERT-based architecture to enhance semantic search and ranking. Additionally, he implemented a proof of concept with Typesense Database to improve search candidate generation and autosuggestion corpus, which resulted in a better click depth ratio. His work also includes establishing an in-house feature store using Kafka, Spark Streaming, and Redis for real-time feature generation.
Previous Experience
Prior to joining Zepto, Vignesh worked at several organizations in various capacities. He served as a Machine Learning Engineer at Willings, Inc. for a brief period in 2021 while working on a project for Otsuka in Tokyo, Japan. He also held the position of Data Engineer at Vahan for four months in 2022 in Bengaluru, Karnataka, India. His experience includes a role as a Software Developer and Open Source Contributor during Google Summer of Code in 2021.
Education and Expertise
Vignesh Venkataraman studied at the Indian Institute of Technology, Roorkee, where he pursued a dual degree in Physics, achieving an Integrated BS + MS from 2017 to 2022. His academic background provides a strong foundation in analytical and technical skills, which he applies in his professional work in machine learning and data engineering.
Technical Contributions
At Zepto, Vignesh has made significant technical contributions, including enhancing delivery time accuracy by 120 seconds and improving retention by 12% through a retrained XGBoost model. He has implemented efficient data and machine learning pipelines using EMR, Glue, and PySpark for inventory time series forecasting. Additionally, he built a retraining pipeline for Iceberg data lake integration and mentored an intern in developing an in-house ETA maps prediction algorithm.