Oct 2023 – Jul 2025 · 1 yr 9 mos

• Developing tooling and algorithms to optimize Large Language Model (LLM) training at massive scale, encompassing thousands of GPUs and TPUs. This is achieved by utilizing an open-source stack including Ray, NVIDIA NeMo, MaxText, and Kubernetes.

Oct 2020 – Sep 2023 · 2 yrs 11 mos

• Developed and deployed a suite of machine learning algorithms to automate the moderation of user-generated content on Google Maps.
• We reduced live abuse by more than 90%.
• This initiative culminated in the conceptualization and prototyping of multimodal Large Language Models (LLMs), resulting in a projected reduction of over 80% in manual moderation costs.

Mar 2020 – Oct 2020 · 7 mos · Bengaluru, Karnataka, India

• I was part of fraud detection and machine learning platform team.
• I worked on edge compute service which is used for directly pushing ML models in edge devices like android and iOS phones
• I designed a online scoring platform that can serve machine learning model predictions in real time

Aug 2015 – Jan 2020 · 4 yrs 5 mos · Bengaluru Area, India

• Designed a LSTM-CNN-CRF based neural network which extracts meaningful information like product type, brand, color etc from title, description and images. We achieved an accuracy of more than 97% across millions of products.
• Improved current spell correction algorithm using probabilistic Levenshtein distance, language modelling and Viterbi algorithm. Spell correction improved by more than 85%
• Proposed a new siamese neural network for generating multi-modal embedding of products from images, text and user behaviour.
• Rewritten and refactored various Solr components to port the entire search infrastructure to a new serving system based on Solr6. The work involved extensive GC fine tuning and JVM heap analysis. The new system is capable of serving more than 2000qps
• Responsible for end to end design and development of scalable and versioned configuration management for the entire company using git, redis and play.

Aug 2013 – Jul 2015 · 1 yr 11 mos · bangalore

• I worked in the intersection of large-scale graph data analysis, machine learning and distributed systems. I published multiple papers in top-tier conference in my area
• I received gold medal for best outgoing MTech student in my batch

Feb 2012 – Jul 2013 · 1 yr 5 mos · Hyderabad Area, India

• In Bravo Lucy i had the opportunity to work on cutting edge sports analytics application. As part of my job I developed many complex visualisations from large scale real time data.