Vinit Jain | I work on robotics and machine learning.

Vinit Jain

profile image

Hi, I'm a MSc graduate in Computer Science from the Technische Universitat Berlin, Germany and KTH Royal Institute of Technology, Sweden. As part of my graduate program, I have a minor in entrepreneurship and innovation because I really believe in solving problems and building products that people want.

I have programming experience with C++ and Python, and extensive experience with frameworks like Tensorflow, PyTorch, and OpenCV. I also have limited experience with computer graphics tools like PyOpenGL, Blender, and MeshLab. For my work, I usually use tools like Git, Docker, Linux, ROS and Latex. My research interests include Machine Learning, Robotics, and Computer Graphics. Currently, I'm looking for full-time opportunities that involve a mix of machine learning and software development.

When I am not working, I usually run, hike, travel, read, and listen to podcasts. I'm on Twitter @vinitjsjain and GitHub @vinittech. You can also drop me an email at vinitjsjain@gmail.com

Projects

  • Denoising SAR Images: Developed and Trained a deep residual CNN pipeline in linear and logspace to eliminate speckling for SAR images. The dataset was obtained from the European Space Agency by its Sentinel-1A satellite. Our model predicted better results ( > 98% accuracy ) with higher processing efficiency (runtime) when compared against the current state-of-the-art SAR filtering methods like Lee, Frost, and BM3D. Binary Neural Networks using BoP optimizer
  • Binary Neural Networks using BoP optimizer: Reimplemented and validated the 2019 NeurIPS paper on Binary Neural networks using a Binary optimizer where the model is trained and optimized with binary values[1,-1]. Further extended the work by tuning thehyperparameters, and trained on different datasets and architectures, thereby evaluating the robustness ofBinary Neural Networks against the conventional CNNs.
  • An initial solution to MAPC 2019: Developed a strategy for making multiple agents work together by submitting as many tasks as possible in a given time-frame. Implemented Path Planner(A-star), Exploration, Dynamic Mapping and Map-merging,Task-Allocation, and Decentralized Agent-to-Agent Communication.
  • Underwater broadband communication: Designed and implemented an acoustic system for underwater communication. Implemented using Maltab and Simulink.
  • more...


    • Follow @vinittech