Aug 2022 – Aug 2024 · 2 yrs · Bangalore Urban, Karnataka, India

• ▪ Led the fraud AI team in designing and building a state-of-the-art real-time
• fraud detection system (e-commerce) and a real-time feature store for the
• models.
• ▪ Built an ensemble of graph neural network and transformer models to
• identify fraud orders (e-commerce - Wrong product returns, RTO etc.). This
• helped in reducing return fraud by 25% and saves millions in refunds for
• the company every month.
• ▪ Aided in hiring and assembling an exceptional ML team. Mentored junior
• colleagues, guiding them through high-level designs and code reviews.
• ▪ Built generic tools that are used across the data science team for model
• performance tracking, feature drift monitoring and logging.
• ▪ Built real-time inference end-points for the model and batch processing
• pipelines that compute the input features for the models at regular
• intervals efficiently using hundreds of millions of data points (Spark).
• ▪ Responsible for maintaining and updating multiple existing ML models.

Nov 2021 – Aug 2022 · 9 mos · Bangalore Urban, Karnataka, India

Apr 2021 – Nov 2021 · 7 mos

Jun 2019 – Apr 2021 · 1 yr 10 mos

• ▪ Built and deployed a predictive model to rank millions of health insurance customers based on their risk of hospitalisation in the near future which helped the company provide efficient care and reduce the health care costs by 30 percent.
• ▪ Used various clustering techniques to identify patterns and groupings
• of high risk health insurance customers.
• ▪ A number of analyses that drove impactful decisions through effective story telling.
• ▪ Helped the company’s transition to cloud based parallel computing
• platforms by converting existing SQL pipelines to spark based databricks pipelines thereby reducing downtimes to near 0 %.
• ▪ Organised all the data sources to create a data store and standardised the data formats, access mechanisms etc. for easy use across the company.

May 2018 – Jul 2018 · 2 mos · Kolkata, West Bengal, India

• Used hypothesis testing and developed a maintenance plan for the
• cigarette filter manufacturing machines based on the historical
• maintenance data of similar machines in other plants and reduced
• downtime by 20 percent.