Aug 2020 – Dec 2021 · 1 yr 4 mos · Gurgaon, Haryana, India

• 1. Formulated and developed a graph-based representation learning algorithm for cards and merchants to detect transaction frauds which showed a considerable performance boost (≈ 6% AUCPR) over existing algorithms and demonstrated a good trade-off in terms of training time vs. performance. The work resulted in a successful research publication.
• 2. Designed and implemented a novel memory-efficient (≈ 80% reduction in memory usage than the state-of-the-art model) tabular GAN architecture for the enhanced Threatscan product by Mastercard. The model is capable of handling datasets with large cardinality. This work resulted in a successful research publication.

May 2019 – Jul 2019 · 2 mos · Bangalore

• Implemented and simulated the MAS5G architecture, a new 5G mobility scheme, published in IEEE FiCloud, 2019 as a part of the Networks team.
• Locally deployed and tested a proof-of-concept version of the architecture using Kubernetes, Cassandra and Node.js.
• Was given a return offer to join as a Research Engineer.

Dec 2018 – Jan 2019 · 1 mo · Kanpur Area, India

• Worked on Linux Malware Detection at the C3i Center at IIT Kanpur. Developed an end-to-end ML pipeline for classifying a given ELF (Linux executable) file as Malware or Benign on the basis of a variety of Static and Dynamic analysis techniques.
• The developed system is now deployed on their internal network.

May 2018 – Jul 2018 · 2 mos · Pune Area, India

• 1. Worked on an OCR + NLP-based model to extract and label segments of text from PDFs. Completely revamped the existing pipeline to make an 80% faster and more accurate (≈ 92%) system.
• 2. Experimented with image processing methods and Faster-RCNN model for detection and extraction of tables from PDFs.