Ranjith Merugu

Ranjith Merugu

ML & Quant Researcher | Full Stack Developer | Competitive Programmer

MSCS Researcher at Stony Brook | Ex-Samsung Research | Computer Vision, Quantitative Analysis, and AI

LinkedIn Scholar Contact Me Play: Shortest Path Visualizer

About

I am a research-driven computer scientist with prior experience at Samsung Research, where I worked on cutting-edge computer vision problems, published papers at top conferences (CVPR, ICCV), and filed patents in video restoration and model compression.

Currently, I am a graduate researcher at Stony Brook University, working with Prof. Dr. Shubham Jain on model merging, hyperspectral image (HSI) restorations (in collaboration with Prof. Xin Yuan), and innovating advanced math and algorithms in machine learning. My research is focused on scalable model merging, self-supervised learning for HSI, and integrating mathematical rigor into ML for real-world impact.

I am passionate about competitive programming (Codeforces, ICPC, Codevita, Leetcode) and actively mentor students as a Bosscoder DSA Mentor. I enjoy solving algorithmic challenges and sharing knowledge with the community.

Google Scholar: My Profile

Experience

Stony Brook University
MSCS - Research Assistant, Stony Brook University Jan 2025 - Present
Google Scholar: My Profile
Advisors: Prof. Shubham Jain
HSI Restorations: Working with Prof. Xin Yuan
  • Research in Multi-Modality and Computer Vision (CV) for unique use cases.
  • First-authored paper on model merging (NeurIPS 2025). StatsMerging
  • IMU pose estimation (WACV/AAAI submission, 2nd author).
  • HSI restoration research (AAAI/CVPR submission).
  • Joint Flow and Feature Refinement for Video Restorations. JFFRA
Tech Stack: Python, PyTorch, TensorFlow, OpenCV, Scikit-learn, Computer Vision, Model Merging, Research, Deep Learning, NLP
Samsung Research
Senior Software Engineer, Samsung R&D Institute India Mar 2024 - Jan 2025
Collaborators: Dr. Jitesh Kumar Singh, Dr. Siddharth Roheda, Dr. Amit Satish Kumar, Pankaj Kumar Bajpai
  • Submitted an Paper as First Author on Video restoration enhancements using Flow and Attentions (Rebuttal CVPR25) (Advisor: Dr. Amit Satish Kumar)
  • Developing a video denoising network specifically for low-light scenarios to enhance visual quality in dark conditions.
  • Working on developing a compressed AI network using multi stage (VQ-VAE) to improve efficiency and reduce computational resources. Current status of computational reduction was 4x. (In collaboration with Dr. Jing Li)
  • Working on an idea of utilizing VQ-VAE codebook as prior to do restoration tasks on Image and Videos, for now worked on POC for denoise and deblur using codebook as prior (Paper work is under progress). (Advisor: Pankaj Kumar Bajpai)
Tech Stack: Python, TensorFlow, Android (NDK), OpenCV, Flask, C++, Bash, Blender, Git, REST APIs
Samsung Research
Software Engineer, Samsung R&D Institute India Jul 2022 - Mar 2024
Collaborators: Dr. Jitesh Kumar Singh, Dr. Siddharth Roheda
  • Worked on development of occlusion and optical flow network, to enhance the quality of under display cameras using multi-frame captured (Under Display Array Cameras). Filed a patent for this project. Work published in CVPR - 2024 paper. (Advisor Dr. Jitesh Singh)
  • Designed and developed an E2E framework to deploy the multiple models on flagship mobiles for the media quality assessment (MDC), the MQA assesses the quality of images and videos and helps to run multiple enhancements. (Received best MDC project award) (Advisor Dr. Siddharth Roheda)
  • Optimized blur detection model on device execution time by 100 ms using multi-threading.
  • UDAC: Under-Display Array Cameras
Tech Stack: Python, PyTorch, OpenCV, Computer Vision, Deep Learning, Optical Flow, Segmentation, Android (NDK), C++, Flask
Infineon Technologies
Infineon Technologies, Bangalore, India
Software Engineer Intern, Received Best Intern Award
Jun 2021 – May 2022
  • Designed and implemented a GUI tool for EA model validation and automated C-code generation.
  • Used by software architects across departments; awarded Best intern project.
Tech Stack: Python, Mako, Flask, REST API, Jira, EA, Jenkins, Architecture Tools
Juspay
Juspay Technologies, Bangalore, India
Software Engineer Intern
Mar 2021 – Jun 2021
  • Improved response time of RESTful APIs by optimizing Redis and PostgreSQL operations.
  • Contributed to RESTful API development and code migration using PureScript/Haskell.
Tech Stack: Haskell, Node.js, Redis, PostgreSQL, Postman

Education

Stony Brook University

MS, Computer Science (Jan 2025 - Dec 2026)
GPA: 4.0/4.0
Courses:
  • CSE 527 - Computer Vision (Dr. Zhaozheng Yin)
  • CSE 548 - Analysis of Algorithms (Dr. Rezaul Chowdhury)
  • CSE 593 - Independent Research (Dr. Shubham Jain)
  • CSE 535 - Distributed Systems (Dr. Mohammad Amiri)
Activities: Competitive Programming and Quant Club Active Member

Lovely Professional University

BTech, Computer Science and Engineering (Jun 2018 - May 2022)
Grade: A+
Achievements: ICPC Nationals, Codevita Rank 4, Facebook HackerCup Round 2
Courses:
  • Artificial Intelligence
  • Machine Learning
  • Data Structures
  • Algorithms
  • Computer Networks
  • Operating Systems
  • Database Systems
  • Formal Languages & Automata
  • Software Engineering
  • Probability & Statistics

Relevant Coursework

ML & AI

  • Artificial Intelligence
  • Machine Learning Algorithms
  • Deep Learning (AI)
  • Computer Vision
  • Natural Language Processing (NLP)
  • Statistical Methods in AI
  • Image Processing Algorithms

Quant & Math

  • Probability & Statistics
  • Stochastic Calculus
  • Optimization
  • Linear Algebra
  • Differential Equations
  • Time-Series Analysis

Dev & Systems

  • Data Structures & Algorithms
  • Operating Systems
  • Computer Networks
  • Database Systems
  • Software Engineering
  • Formal Languages & Automata

Projects

SS-HSI: Self-Supervised Learning for Hyperspectral Image Reconstruction [AAAI 26]

Unified self-supervised learning framework for hyperspectral image reconstruction via mask guidance. AAAI Under Review.

Tech Stack: Python, Deep Learning, Hyperspectral Imaging, Self-Supervised Learning

StatsMerging: Statistics-Guided Model Merging [NeurIPS 25] [arXiv]

Developed a novel model merging technique for task-specific teacher distillation. Submitted to NeurIPS 2025.

Tech Stack: Python, PyTorch, Model Merging, Knowledge Distillation, SVD, Deep Learning

Joint Flow And Feature Refinement For Video Restorations (JFFRA) [ICCV 25] [arXiv]

First-authored paper on model merging for video restoration using attention and flow, submitted to ICCV/CVPR/NeurIPS.

Tech Stack: Python, PyTorch, Deep Learning, Attention, Video Restoration, Model Merging

CDVS: 8K Video Compression using Multi-Stage VQ-VAE [CVPRW 25] [CVPRW Link]

Developed a memory-efficient 8K video compression approach using a multi-stage VQ-VAE architecture, achieving 4x reduction in computational resources. Designed and implemented the pipeline for on-device slow-motion video generation. Filed a patent for this work.

Tech Stack: Python, TensorFlow, VQ-VAE, Video Compression, Deep Learning, Patent Filing

Under-Display Array Cameras (Samsung Research America) [CVPR 24]

Developed a distributed vision pipeline combining occlusion-aware optical flow and segmentation with multi-scale feature fusion. Delivered a low-latency, high-resolution (30ms UHD) system for real-time under-display camera image enhancement. Impact: AI patent filed, paper in progress.

Tech Stack: Python, PyTorch, OpenCV, Computer Vision, Deep Learning, Optical Flow, Segmentation

Media Quality Assessment Framework (SRIB)

Developed an IQA module to evaluate image/video restoration models, focusing on human perception sensitivity. Received MDC Award.

Tech Stack: Python, Flask, OpenCV, Image Quality Assessment, Video Processing

Sentiment Analysis - RNN

Developed a sentiment analysis model using RNNs on financial/news data, achieving 83% accuracy on Kaggle dataset.

Tech Stack: Python, TensorFlow, NLP, RNN, Sentiment Analysis

Programming & Competitive Coding

Competitive Programming Achievements

  • 3500+ problems solved across Leetcode, InterviewBit, GeeksforGeeks, SPOJ, Codeforces (Expert, 1691), Codechef (5-Star)
  • Qualified Facebook HackerCup Round 2, TCS CodeVita Rank 4, ICPC-Regionals
  • Ranked #1 in Stony Brook weekly contests (peak rating 1900)

Algorithmic & Coding Skills

  • Proficient in advanced algorithms: FFT, NTT, InverseFFT, Generating Functions, Binomial Heaps, Deterministic Algorithms, Binary Lifting, DSU on Tree, HLD, Centroid Decomposition, Square Root Decomposition, DP (e.g., Divide and Conquer, Knuth, Li Chao Tree)

Mentorship

  • Bosscoder DSA Mentor (2022-now): Mentored 100+ students, 80+ placed in top product-based companies

Current Research & Algorithmic Innovation

Restorations on Hyperspectral Imaging & Model Merging in ML

  • Extending model merging for unique ML applications, including hyperspectral image restoration and real-time video analytics.
  • Innovating by integrating advanced math and programming algorithms into ML, replacing traditional approaches. Targeting top conferences: CVPR, AAAI, NSDI.

Current Submissions

  • NSDI 26: Model Merging for Real-Time Video Analytics
  • AAAI (Under Review): SS-HSI: A Unified Self-Supervised Learning Framework for Hyperspectral Image Reconstruction via Mask Guidance

Programming Passion & Mentorship

  • Active problem solver on Codeforces, SPOJ, Atcoder, Leetcode (Ranjith123), coding since 2019.
  • Bosscoder DSA Mentor (2022-now): Mentored 100+ students, 80+ placed in top product-based companies.

Publications

Joint Flow And Feature Refinement Using Attentions For Video Restorations [arXiv]

ICCV 25 (Rebuttal)

StatsMerging: Statistics-Guided Model Merging via Task-Specific Teacher Distillation [arXiv]

NeurIPS Submitted

UDAC: Under-Display Array Cameras

2023 - 2024 (Advisor: Jitesh)

CDVS: Compressed Domain On Device Memory Efficient 8K Video SlowMo

CVPRW 2025

Image Multi Enhancement for Under Display Cameras

Patent Filed

Skills

Programming: Java, C++, Haskell, Python, TypeScript
ML/AI: Computer Vision, NLP, Deep Learning, Model Compression, Quantitative Analysis
Frameworks: Spring Boot, React, Node.js, Flask, Django
Tools: AWS, REST APIs, Git, Docker
Competitive Programming: 5+ years, DSA Mentor, Leetcode: Ranjith123

Recommendations

Khadija Tahseen (Software Engineer, Ex-Amazon, JPMorgan):
"Deeply thankful to someone who quietly but consistently supported me during my preparation journey. The mock interviews, honest feedback, and clarity in guidance made a real difference. Sometimes, the people who stay behind the scenes have the biggest impact — this was one of those cases."

Contact

Email: meruguranjith13@gmail.com
LinkedIn: linkedin.com/in/ranjith-merugu
Leetcode: Ranjith123
Location: Stony Brook, NY, USA