Jul 2024 – Jul 2024 · 0 mo · Belgrade, Serbia · On-site

• Attended the prestigious EEML (Eastern European Machine Learning) Summer School in Serbia, where I had the opportunity to engage with leading experts in the ML/DL domain. Gained invaluable insights from their expertise while exploring advanced machine learning techniques, fostering research collaborations, and exchanging ideas with a global community of like-minded professionals. I also had the opportunity to present a poster. Received the travel grant from Google to attend the summer school.

May 2022 – Oct 2025 · 3 yrs 5 mos · Gurugram, Haryana, India

• 1. Designed and developed a novel Graph Neural Network model for Fraud Detection (MRP-GNN), consolidating carding information from multiple sources. This project is live and results in 20% increase in the detection of fraudulent merchants.
• 2. MRP-GNN was able to flag 5,862 high-risk merchants and identify $17M of additional fraud in Europe that went undetected by other processes.
• 3. Mentored interns to implement a novel neighbor sampling technique, significantly improving the efficiency of the fraud detection model.
• 4. Working on further extending the MRP models to predict the risky merchants.
• I, along with other scientists at Mastercard and professors from the Indian Institute of Technology, Delhi (IIT-D), co-organized a local meetup for the Learning on Graphs (LoG) Conference 2024.

May 2021 – May 2022 · 1 yr · Bengaluru, Karnataka, India · Hybrid

• 1. Worked on Multi-label automatic categorization problem. The aim was to perform data driven classification of AZURE offers.
• 2. Experimented various ways to embed textual attributes of the offers in low-dimensional space (e.g., BERT, Word2vec, FastText, Tf-Idf, etc).
• 3. Also, tried various models for multi-label text classification, including logistic regression, SVM, tree-based models, and various neural network architectures.
• 4. Built a crawler for Azure websites to collect data on all partners and their published offers due to the unavailability of ground truth data.
• 5. Proposed a clustering model combining PCA, UMap, and KMeans for unsupervised categorization without relying on existing category information during the model-building phase.

Sep 2020 – May 2022 · 1 yr 8 mos · Bengaluru, Karnataka, India

• 1. Worked on the Pricing Portfolio Project to automate style prices (discounts) at Myntra. Explored various approaches to feature engineering and tested different models for demand forecasting.
• 2. Also, proposed a novel graph neural network-based approach for User Cohort Pricing.
• 3. Worked on a recommender system, leveraging the power of network embeddings through Graph Neural Networks.
• 4. HackerRamp, Myntra: 1st runner-up of HackerRamp (among 30+ teams) organized by Myntra.

Feb 2018 – Mar 2018 · 1 mo · Metcalfe House, Delhi, India

• I was fortunate enough to be selected for the JRF post at CEPTAM, Metcalfe House, DRDO where I got opportunity to work with highly talented scientists.