Epic Systems - Verona, WI

Software Developer - MyChart

Apple Inc. - Cupertino, CA

Data Engineering Intern - Apple Maps

  • Developed a data processing (ETL) pipeline using Pyspark & Hadoop to ingest new modality data into ML model for NextGen destination point prediction, achieving ~20% improvement in validation metrics.
  • Revamped image-segmentation models from the production code of two more ML applications in Maps.
  • Improved bike-lane prediction metrics from 78% to 91% and road-segmentation metrics from 85% to 92%.

Carnegie Mellon University - Robotics Institute, PA

Apple Device Recycling Project - R&D Intern - Biorobotics Lab

  • Collaborated with Apple Inc. to revolutionize its recycling robots, Daisy and Dave, using machine learning.
  • Prototyped a novel multimodal attentional CNN (iCAM) for the intraclass iPhone classification task from images and X-ray scans, achieving 99.9% accuracy and improving by 14% over the state-of-the-art.
  • Paper accepted for publication at IEEE International Conference on Intelligent Robots and Systems (IROS), 2022. [CMU webpage]

Point Cloud Registration Project - Research Intern - Biorobotics Lab | [link] May 2020-Aug 2021

  • Designed a correspondence-based novel loss function for the 3D point cloud registration task, achieving a decrease of 40% in rotation error over the state-of-the-art loss functions.
  • Developed an outlier filtering approach that decreased registration error by 80% over unfiltered point clouds.
  • Published at IEEE International Conference on 3D Vision (3DV), 2020. [Paper][IEEE]

Autonomous Driving (AD) Simulator Project - R&D Intern - RPAD Lab | [link] May 2021-Aug 2021

  • Programmable Light Curtains have potential to replace the LiDAR sensor, cutting the AD costs by upto 80,000$.
  • Designed a Light Curtain simulator & integrated it with an AD agent achieving a 7.2% increase in driving score. [ppt]

ESIEE Paris

Research Intern at LIGM Laboratory, ESIEE Paris (Mentor: Prof. Laurent NAJMAN)

  • Honored to receive the prestigious CHARPAK Scholarship by the French Government to fund this research.
  • Project Title: “Learning Representation using Mathematical Morphology.”
  • Analysed Watershed Cuts, a Graph-based Segmentation approach, for Semi-Supervised Classification.
  • Integrated Deep Learning Models with Classical Watershed Cuts to Improve (~5%) the Classification Results. [Report]

Indian Institute of Technology Madras, Chennai, India

3D Semantic Segmentation Project - Student Researcher - Computer Vision Lab | [link] Nov 2019-May 2020

  • Independently proposed a novel 3D deep learning architecture for unstructured data such as 3D point clouds.
  • Achieved state-of-the-art accuracy (93.9%) on the ModelNet40 3D object classification task. [Paper]

Deep Learning based RANS Method for Turbulence Modeling - Undergraduate Research - Aerospace Department, IIT Madras

  • Developed a Tensor Basis Neural Network model to predict the Reynolds Stress Tensor for Turbulent Flows.
  • Achieved a 2.8% RMSE between True & Predicted Values across Flows with different Reynolds Numbers. [Report]