Education
- M.S. in Computer Science and Engineering, University at Buffalo, 2024-26
- Grade: 3.95/4.0
- Research Track with specialization in Computer Vision and Deep Learning
- Courses: Deep Learning, Reinforcement Learning, Computer Vision, Machine Learning, Parallel and Distributed Processing, Information Retrieval, Algs for Modern Computing Systems
- B.Tech. in Electronics and Communication Engineering, Vellore Institute of Technology, 2017-21
- Grade: 9.48/10.0
- Courses: Applied Linear Algebra, Computer Communication Networks, Data Structures and Algorithms, Digital Image Processing, Embedded System Design, Probability Theory and Random Processes, Statistics for Engineers
Work experience
- ML Computing Intern, May 2025 - Aug 2025
- Mercedes-Benz R&D (MBRDNA), San Jose, CA
- Duties included: Autonomous Driving
- Benchmarked TensorRT and authored a report for next-gen Autonomous Driving compiler stack.
- Optimized inference by exploring CUDA Graphs with Nsight Systems, trimming CPU launch overhead and cutting latency 20%. Drove 4× faster inference via INT8 implicit and explicit quantization.
- Built a C++/CUDA utility to inspect TensorRT engine weights and automate network-pruning sweeps.
- Implemented unified-memory zero-copy paths and custom GPU pools, lowering peak GPU use 18% and eliminating two memcpy calls per frame.
- Proved bit-exact determinism at runtime & layer level; stress-tested execution contexts to meet ASIL-D safety.
- Research Assistant, Oct 2024 - Present
- The Research Foundation for SUNY, Buffalo, NY
- Duties included:
- Developed a high-density object counting solution using GroundingDINO integrated with Swin-B and BERT encoders.
- Implemented a feature fusion module with bidirectional cross-attention across text and image modalities, enhancing model interaction.
- Improved performance through semantic-guided visual and textual prompt tuning combined with domain-specific loss functions, achieving a Mean Absolute Error (MAE) of less than 1.
- Supervisor: Dr. Junsong Yuan, Dr. Peter CW Kim
- Graduate Student Assistant - Grader, Jan 2025 - May 2025
- University at Buffalo, Buffalo, NY
- Duties included: Course Grader for ‘CSE574 Intro Machine Learning’, evaluating student quizzes and assignments
- Supervisor: Dr. Asif Imran
- Software Engineer - II, July 2021 - July 2024
- Citi (CSIPL, Citigroup), Chennai, India
- Duties included:
- Co-owned and led the development of a web-based end-to-end tool initiative for data creation and conditioning, reducing the time and effort required for data handling. Engineered the tool using ReactJS and NodeJS, optimizing test data creation and conditioning workflows
- Developed API layers in .NET using minimal-API, exposing Selenium-C# automation functionalities and enabling remote headless execution which decreased data creation time by 5x, reduced cross-team dependency and turnaround time by 80%
- Designed C# automation scripts for Fund Transfer module, increasing the test coverage for 7 market countries by 50% and saving cost in licensing tools by 30%
- Computer Vision Intern, Jan 2021 - June 2024
- PlaEye LLC
- Duties included:
- Logo Detection and Analysis in Images: Created a brand analytics system and an Android application that can detect, and classify different sports brands in a given image scene captured from the app and list down the number of instances for each brand. The model was trained on LogoDet-3k dataset and the inference model was run using ArmNN for arm64-v8a architecture devices
- Dealt with the challenging in-the-wild dataset marked by class imbalance and diverse image features, including textual and visual logos, with scaled loss functions and class augmentations
- Computer Vision Intern, Apr 2020 - May 2020
- PlaEye LLC
- Duties included:
- Developed a novel algorithm for automated tennis court calibration using Homography and Cross ratios as the primary basis to address the challenging aspect of single-view calibration systems where it is difficult to extract the feature points from partially occluded or poorly recorded court area images/video feeds
- Created UI-based tools (using PyQt) for perspective projection and camera calibration
- Contributed to optimizing the background subtraction algorithm SubSENSE for GPU using CUDA with C
- Helped in creating a depth estimation algorithm from 2D-to-3D translation with the help of perspective projection and shadow point computation using ray geometry which was later used by the team to determine the position and spin of tennis balls from the video feed
Patents and Publications
- Multi-Agent Computer Vision system & methods for automated object counting, classification and tracking. USPTO Application No. 63/851,812. Status: Patent Pending.
Honors & Awards
- Best AI/ML Project, UBHacking’24, Nov 2024
- Won the ‘Best AI/ML Project’ award at UBHacking’24 for developing an AI-powered, context-aware bookmark manager extension for Chrome
Awarded a top rating (1 out of 4) for innovation in developing automated reporting tools and dashboards at Citi
- Academic Meritorious Award: Achieved 4th overall department rank in B.Tech Electronics and Communication Engineering, 2017-21 batch