PhD Student in Computer Science
CUNY Graduate Center
I am an incoming PhD student in Computer Science at the Graduate Center, City University of New York (CUNY), where I will be advised by Prof. Raffi Khatchadourian. My research interests include software engineering, program analysis, programming languages, and machine learning systems.
I received my Master of Science in Computer Science and Engineering from the University at Buffalo and my Bachelor of Technology in Electrical and Electronics Engineering from Visvesvaraya National Institute of Technology, Nagpur. During my undergraduate studies, I was advised by Dr. S. S. Bhat, with whom I worked on a SPARC-funded Global Grid research project.
You can reach me at: vtarungangadhar[AT]gmail[DOT]com
At Hunter College, I work on static program analysis for machine learning systems. My research focuses on understanding how ML programs use frameworks and APIs, identifying correctness and maintainability issues, and developing automated analysis and transformation techniques to improve software reliability and quality.
In my final year undergrad I worked on a SPARC-funded Global Grid project. My work involved developing a Python-based compiler and parser to speed up model processing and integrating multiple solver APIs including Gurobi and CPLEX. With optimization techniques in NumPy, SciPy, and multiprocessing, solver runtimes were improved by more than 25 percent.
At Hitachi I worked with Microsoft Azure to improve data pipelines for Canadian projects. I analyzed SQL control tables and added over fifty new data points that helped the sales team get better insights. I also set up Azure Data Factory and Logic Apps workflows along with a CI/CD pipeline that made recurring data operations faster and easier to manage.
At Verticross India, I contributed to enterprise software solutions for the power, oil and gas, and environmental sectors. I worked with Java and SOAP-based web services to develop and integrate backend components, enable data exchange between applications, and support existing software systems. I also assisted with testing, debugging, and maintaining services used in industrial applications.
Phys-GS Lite is a lightweight prototype that augments 3D Gaussian Splatting with semantic separation and physics-inspired dynamics. Using simple geometry-driven heuristics, the method distinguishes between rigid and deformable regions . A sinusoidal wind field introduces wave-like deformation, while graph Laplacian smoothing ensures coherent motion.
Enhanced 3D Gaussian Splatting with semantic awareness by fusing 2D SAM segmentations through multi-view voting, producing semantically labeled splats for object-aware rendering and editing on the Playroom dataset without requiring 3D annotations.
This project explores unpaired image-to-image translation for converting daytime driving scenes into nighttime equivalents using CycleGAN. Unlike paired supervised methods, CycleGAN leverages cycle-consistency and adversarial training to learn mappings between Day ↔ Night domains without requiring aligned image pairs.
Developed a data-driven framework to identify optimal urban tree planting sites in Buffalo, NY using over 133,000 records from the Bureau of Forestry dataset. Applied clustering, regression, and decision tree models alongside geospatial visualization to uncover eco-benefit gaps and highlight priority zones. Produced interactive heatmaps and maps that provide actionable insights for sustainable urban forestry planning.
Developed optimization algorithms for microgrids and interconnected global grids using the Graph Based Optimization Modeling Language (GBOML) in Python. Modeled solar, battery, and demand nodes, integrating multiple solvers including Pyomo, PuLP, SciPy, and Gurobi to evaluate performance.
Developed a machine learning framework for electricity consumption forecasting using 12 years of quarter-wise campus data, applying linear and polynomial regression models after rigorous data cleaning and feature engineering.
Built a Streamlit web application to compute and visualize zero-input, impulse, step, and total responses of first- and second-order LTI systems, integrating SymPy for symbolic derivation and NumPy/Matplotlib for simulation with deployment on Hugging Face Spaces.