Projects

A portfolio of research projects, applied AI solutions, and technical innovations spanning computer vision, robotics, and machine learning systems.

5
Total Projects
1
Active Projects
1
In Production
2
Research Projects

Open Source Contributions

Kornia

🐍 Geometric Computer Vision Library for Spatial AI.

⭐ 10.6k starsView on GitHub →

ReinforceBots

Developing a chatbot by combining RL and NLP concepts.

⭐ 3 starsView on GitHub →

Safety driving assist

Driver assist system to warn/stop the vehicle based on alertness score, calculated based on facial expressions.

⭐ 9 starsView on GitHub →
View All Contributions
Category:
Status:

Showing 5 of 5 projects

FitDit Vision Transformer Implementation

Research2024-2025
Completed
🔬
Project Screenshot

Novel approach to convert flat-lay to model images with FitDiT architecture, refined apparel masks to improve garment alignment and realism, attention mechanism improvements, and training strategies.

Technologies Used

PyTorch,Transformers,CUDA,TensorBoard,Weights & Biases,

Key Objectives

  • •Increased apparel context by change in masking
  • •Updated loss function to use DFT loss for better convergence and fidelity

Project Impact

Better texture
Quality Gain

Collaborators

NeuroPixel.AI Research,

Real-time Segmentation Pipeline

Applied AI2024 - Present
Production
🤖
Project Screenshot

Production-grade object segmentation system optimized for real-time performance in apparel segmentation application, featuring TensorRT conversion and efficient post-processing.

Technologies Used

Segment Anything Model (Meta AI),TensorRT,CUDA,Docker,Kubernetes,FastAPI,

Key Objectives

  • •Process 0.5 FPS on standard CPUs
  • •Deploy scalable inference services
  • •Implement model versioning and A/B testing

Project Impact

0.5 FPS average
Processing Speed

Collaborators

NeuroPixel.AI,

Active Perception using Light Curtains

Research2023
Collaboration
🔬
Project Screenshot

Collaborative research with CMU RPAD Labs on developing novel perception systems for autonomous vehicles using light curtain sensors for efficient obstacle detection and navigation.

Technologies Used

Python,OpenCV,Light Curtain Sensors,C++,Graph Neural Networks,

Key Objectives

  • •Develop real-time light curtain processing algorithms
  • •Implement adaptive sensing strategies
  • •Create 3D environment reconstruction methods

Project Impact

95% precision
Detection Accuracy

Collaborators

CMU RPAD Labs,Siddharth Ancha,Prof. David Held,
👨‍💻 Code📄 Paper📹 Video

MLOps Platform for Model Deployment

Infrastructure2024 - Present
Active
⚙️
Project Screenshot

Comprehensive MLOps platform enabling seamless model development, validation, and deployment with automated CI/CD pipelines and monitoring capabilities.

Technologies Used

Kubernetes,Docker,Prometheus,Grafana,Loki,GPU-slicing,

Key Objectives

  • •Implement comprehensive model monitoring
  • •Enable A/B testing for model deployments
  • •Create reusable deployment templates

Project Impact

75% faster
Inference Speed

Collaborators

NeuroPixel.AI,DevOps Team,

Intelligent Ground Vehicle Prototype

Robotics2019
Completed
🚗
Project Screenshot

Design and development of an autonomous ground vehicle capable of navigation in constrained environments using sensor fusion, path planning, and lane detection algorithms.

Technologies Used

ROS,Gazebo,OpenCV,Python,Arduino,Sensors,

Key Objectives

  • •Implement autonomous navigation in GPS-denied environments
  • •Develop robust obstacle avoidance algorithms
  • •Create sensor fusion for localization

Project Impact

Published in ICCRT 2019
Navigation Accuracy

Collaborators

Autonomous Ground Vehicle Research Group, IIT Kharagpur,
📄 Paper

Interested in Collaborating?

I'm always excited to work on innovative projects that push the boundaries of AI and create real-world impact. Let's build something amazing together.

Start a Project