Machine Learning for Ads/Product Retrieval across Facebook and Instagram (graphs, retrieval, deep two tower networks, ranking, generative retrieval)

Primarily working on two major areas:

Generative AI Model Engineering - Building LLM based code completion systems (C++, MATLAB)

AI Services - Building the infrastructure for Generative AI applications (Go, C++)

AI Assisted Coding

• Worked on a transformer based architecture for code prediction that is fast, accurate, runs fully offline, and meets the RAIL guidelines for speed

• Reduced memory footprint by 53% and sped up the model inference by 17% using network projection techniques

• Increased model accuracy by 12.3% by incorporating improved data preprocessing techniques using Abstract Syntax Trees (AST) and Byte Pair Encoding (BPE)

• Data Analysis for tracking user productivity metrics like acceptance rate, correctness, etc.

Conferences, Workshops, Talks

• NeurIPS 2023 - Women in Machine Learning: [Poster] Farm to Plate AI

• Google DevFest Boston, December 2023: [Talk] Code Generation and Code Completion using Generative AI

• Women in Data Science, November 2023

• Grace Hopper Celebration 2023: [Workshop] Farm to Plate AI: Enhance Freshness and Reduce Waste using Computer Vision and Robotics

• NeurIPS 2021 - Women in Machine Learning: [Poster] Getting Started with Model Cards

• NeurIPS 2021: [Social] Reducing Maternal Mortality using AI

Rotational Program working with different teams:

• Adding binary feature support for ORB-SLAM workflows, Computer Vision

• Building accuracy and loss progress plotters for Object Detector training (YOLO, SSD, FasterRCNN, RCNN etc.)

• Building datasets from scratch for MATLAB Code for training AI-Assisted Code Generation models

Teaching Assistant at CMU's School of Computer Science for the course Mobile & IoT Computing