cv

Master's in Computer Science

Stony Brook University, NY, USA

• Advanced Project - Multi View Gaze Following
• Relevant Coursework
  • Computer Vision.
  • Big Data Analytics.
  • Natural Language Processing
  • Operating Systems

Bachelor's in Electrical and Electronics Engineering

Birla Institute of Technology and Science, Pilani, Rajasthan, India

• Relevant Coursework
  • Neural Networks and Fuzzy Logic
  • Data Structures and Algorithms
  • Object Oriented Programming
  • Operating Systems

Computer Vision Intern

Zebra Technologies, Holtsville, NY

• Working on optimizing and deploying Vision transformer models for semantic segmentation task on edge devices.

Research Assistant

Computer Vision Lab, Stony Brook University, NY.

• Working on the Gaze Following task focused on leveraging multiple views to improve current SOTA models.
• Building a highly accurate Multi-view Gaze Dataset with 37k pairs of views at various indoor locations.

Software Engineer

Qualcomm, Hyderabad, India

• Automated fixing kernel crashes by employing web mining which reduced kernel upgrade efforts by over 30%.
• Designed an upstreamable Linux kernel device driver and user space application for seamless IPC over SPI.
• Modified various transport drivers in accordance to Linux RPMSG and suspend frameworks to support Deep Sleep state critical in optimizing power consumption for upcoming chipsets in the smart wearables segment.
• Led a team of 3 for Snap development of user space entities for an Ubuntu Core IIoT platform.

Data Science Intern

Piramal Enterprises Limited, Mumbai, India

• Developed a Retailer Analytics Rule Engine for Piramal's INR 400 million business of OTC and Baby products.
• Profiled 0.25 million retailers into 7 categories based on 5 metrics using Gaussian Mixture Mean clustering.
• Augmented sales by 20% through a Google Maps web crawler creating a database of 1000+ untapped stores.
• Designed a Google News web extractor for the NBFC business to strategize approval and revival of loans.

Surgical Skill Prediction using Unsupervised Tool Segmentation

• Improved segmentation of tools by modifying existing cues and using optical flow for the JIGSAWS dataset.
• Surpassed skill prediction accuracy of the existing model by 2%.

Human-Aware Recurrent Transformer

• Reinstated the importance of user-state by achieving 4% higher accuracy than baseline GPT-2 for hate speech detection.
• Visualized the user-states to prove that no concrete groupings of users exists against their labels for downstream tasks.

Cognitive Prostheses for Goal Achievement

• Implemented the research Paper - Cognitive Prostheses for Goal Achievement by Dr. Falk Lieder.
• Studied and designed a MDP model to compute optimal incentives for to-do list gamification.
• Proposed improvements to existing model by incorporating dynamic priorities in classification of tasks.

Text to Image Synthesis using StackGAN

• Analysed the impact of StackGAN on the image quality for two datasets Caltech-UCSD Birds and Oxford-102.
• Employed Conditioning Data Augmentation to smoothen latent data and improved accuracy using ADAM optimization.

• Rising Star Award
  • Awarded to top 1% of fresh hires at Qualcomm

• Kishore Vaigyanik Protsahan Yojana (KVPY) Fellowship Award
  • All India Rank - 1300