Vikram Kalabi
About Vikram Kalabi
Vikram Kalabi is a Staff Software Engineer currently working at MetaMap in Bengaluru, India. He has extensive experience in data science and engineering, having held various roles at companies such as Monsanto, Tata Consultancy Services, and Inkers Technology.
Work at MetaMap
Vikram Kalabi has been employed as a Staff Software Engineer at MetaMap since 2022. In this role, he has focused on enhancing document processing systems through advanced AI technologies. His contributions include reducing manual review requirements by 73% and decreasing document reading errors by a factor of four using multimodal large language models (LLMs) and robust AI-based validation checks. He has also led the migration to the Triton inference server, resulting in significant improvements in inference speed, ranging from 3x to 50x.
Previous Experience
Vikram Kalabi has a diverse background in software engineering and data science. He worked at Monsanto Company as a Data Scientist from 2015 to 2017, where he was based in Bengaluru, India. Prior to that, he served as a Systems Engineer at Tata Consultancy Services from 2011 to 2013. He also held the position of Analytics Engineer at Impetus Infotech India Pvt Ltd from 2013 to 2015. After his tenure at Monsanto, he co-founded Datalore Labs Private Limited from 2017 to 2019 and later worked at Inkers Technology as Head of AI and Lead Machine Learning Engineer from 2017 to 2022.
Education and Expertise
Vikram Kalabi earned a Bachelor of Engineering (B.E.) in Computer Science and Engineering from Visvesvaraya Technological University, completing his studies from 2007 to 2011. He furthered his education by obtaining a Self-Driving Car Engineer Nanodegree from Udacity between 2017 and 2018. His academic background provides a strong foundation for his expertise in machine learning, computer vision, and AI technologies.
Achievements in AI and Automation
Throughout his career, Vikram Kalabi has achieved notable advancements in AI and automation. He directed the development of a multimodal active learning system that significantly reduces the need for manual data labeling. His work has led to an automation rate exceeding 95% in template matching, which streamlined custom document reading by a factor of 20. Additionally, he added over 100 new templates to enhance the versatility of document processing systems.