Chahat Kalsi

MSCS student at Stony Brook University, NY

About Me

Hello, World!

I'm Chahat Kalsi, a second-year Master's student in Computer Science at Stony Brook University, NY, US.

At the Center of Visual Computing (CVC), I work under Prof. Arie E. Kaufman and Saeed Boorboor, focusing on immersive systems like Silo, a cylindrical stereo display with 168 high-resolution panels. I helped develop SiloEngine, an OpenGL-based framework enabling seamless stereoscopic rendering and interaction. I also created a conformal mapping-based visualization pipeline to address missing floor and ceiling information, collaborating with Prof. Xianfeng Gu on conformal and optimal transport mappings. Currently, I am building a VTK-based volume rendering engine for Silo, with a D3.js iPad interface for real-time interaction.

As a Graduate Assistant at the Center of Excellence in Wireless and Information Technology (CEWIT), I contribute to a vegetation management project, designing advanced deep learning models for detecting vegetation encroachment near powerlines using drone imagery. My work includes developing distributed segmentation frameworks with architectures like U-Net, DeepLabV3+, and PSPNet, achieving state-of-the-art results, including an IoU of 0.70 on the TTPLA dataset.

Current interests:

Publications

Research Projects

Conformal Mapping
Conformal Mapping to Visualize Missing Information in Immersive Systems

Developed an OpenGL based algorithm to render scenes based on input conformal mappings. Deployed it to Silo to visualize its missing floor and ceiling gaps.

SiloEngine
SiloEngine

SiloEngine is an OpenGL framework custom made to be deployed to stereo large-scale immersive systems, featuring stereo or mono rendering modes, gamepad based interaction, and network synchronization using VRPN.

Volume Renderer
Immersive Volume Renderer

This VTK based volume rendering engine is meant to be deployed to large-scale immersive systems, and comes with an iPad deployable d3.js GUI to interact with the volume.

GitHub

Visualization Projects

PokeViz Dashboard
Astronomy Data Visualization Dashboard

A D3.js dashboard for visualizing 5000 years worth of solar eclipses data from NASA. Includes Choropleth map, time series plots, stacked radial bar plots, MDS attributes similarity plots, among others.

GitHub Demo

Learn More
Stereo Rendering
Multidimensional Data Visualization Dashboard

An interactive D3.js dashboard specialized for visualizing multidimensional data, including MDS data and attribute similarity plots and a reorderable and smartly preordered attributes PCP plot.

GitHub Demo

D3 Dashboard
Dimensionality Reduction Tasks Dashboard

D3.js Dashboard with a flask backend for creating visualizations useful for dimensionality reduction tasks using the PIMA Indians Diabetes Dataset. Deployed on an AWS EC2 instance.

GitHub Demo

Experience

Teaching Assistant
Stony Brook University

  • Teaching assistant for CSE 528 (Graduate-Level Computer Graphics) with over 30 students.
  • Guided students in debugging C++ and OpenGL code and understanding graphics-related mathematics.
  • Wrote thorough unit tests using Mockito and JUnit.
  • Graded assignments and exams, managed attendance, and supported course administration.

Aug 2024 - Present

Research Assistant
Center for Excellence in Wireless and Information Technology

  • Designed a distributed PyTorch-based segmentation framework for real-time powerlines detection in drone imagery, leveraging and modifying advanced architectures like U-Net, DeepLabV3, DeepLabV3+, MANet, LinkNet, PSPNet, PAN, and FPN.
  • Integrated attention mechanisms, rich convolutional features, and object association methods into these models, significantly improving model performance, achieving an IoU of 0.70 on the TTPLA dataset.
  • Conducted comprehensive hyperparameter tuning using RayTune and ablation studies to evaluate model performance across different architectures, ensuring optimal performance for real-time applications.
  • Enhanced Fast-SCNN with adaptive anti-aliasing techniques to improve real-time detection accuracy
  • Refined model accuracy through iterative fine-tuning and annotation on a custom drone image dataset.

Sept 2024 - Present

Software Development Intern
Goldman Sachs

  • Automated firm-wide Cash Gap report generation, using Apache Spark in Java for processing data and performing calculations on it, and Apache Kafka for robust data streaming throughout the system.
  • Developed a Java Spring Boot MVC backend to facilitate the Cash Gap report’s approval process.
  • Wrote thorough unit tests using Mockito and JUnit.
  • Practiced agile development methodologies with Confluence and Jira for streamlined collaboration and project management

Feb 2023 - Jun 2023

Summer Analyst
Goldman Sachs

  • Designed pipelines for seamless integration of static CSV data into the system, developing a Java Spring Boot application to persist data in the database and crafting an Apache Kafka publisher for dynamic streaming of this data within the system, while working on Linux environments, followed by thorough unit and integration testing.
  • Enforced adherence to Software Development Life Cycle (SDLC) standards throughout the project, accompanied by meticulous documentation.

July 2022 - Sep 2022

Research Intern
Nottingham Trent University

  • Coordinated and contributed to the research project “Novel Computational Tools to Assess ADHD in Young Adults”, utilizing MATLAB to denoise EEG signals, employing techniques such as ICA, bandpass and notch filtering, and, wavelet transforms.
  • Developed EEG signal visualisation tools in Python using Matplotlib, Seaborn, Pandas and OpenCV.

July 2021 - March 2022

Contact