Aditya Wagh

šŸ‘‹ Hi, Iā€™m Aditya Wagh

Iā€™m a Research Engineer working on Machine Learning + 3D Computer Vision and Robotics.

Following is the list of problems I:

  • Have worked on: Visual SLAM, Visual-Inertial Odometry, Point Cloud Registration, Semantic Segmentation, 3D Object Detection, Machine Learning Inference
  • Want to work on: LLMs/VLMs + Robot Manipulation/Navigation (a.k.a VLAs), Language-Grounded Semantic SLAM, Robot Learning in general.

And this is the list of domains I find exciting:

  • Augmented and Extended Reality
  • Multi-Modal Autonomous Robots - Perception, Localization, Mapping
  • General Purpose Robots - Humanoids, Mobile and Fixed Manipulators,
  • AI Agents (for example - coding agents like Claude Code)

I hold a MS in Electrical Engineering from New York University (NYU) and a B.Eng (Honors) Electrical Engineering from BITS Pilani.

Always eager to learn new things, discuss and collaborate. Feel free to reach out to me via: [LinkedIn][linkedin], [Email][email] or [X/Twitter (Open DMs)][x]

Techstack
Experienced
Learning

Projects

Below is a list of my notable projects. Feel free to have a look at whatever you find interesting!

SuperSLAM
SuperSLAM

SuperSLAM: Open Source Framework for Deep Learning based Visual SLAM

View Code
Deep Image Matching using Local Feature Transformer
Deep Image Matching using Local Feature Transformer

Pose estimation pipeline to enable 3D reconstruction using Local Feature Transformer (LoFTR)

View Code
State Estimation of a Drone by Visual-Inertial Sensor Fusion
State Estimation of a Drone by Visual-Inertial Sensor Fusion

Optimal state estimation using extended & unscented kalman filter implemented using MATLAB

View Code
3D Object Detection in LiDAR Point Clouds
3D Object Detection in LiDAR Point Clouds

Benchmarking of LiDAR 3D Object Detection networks (VoxelNet, PointNet++, PointPillars) on KITTI Dataset.

View Code
Dimensionality Reduction using Convolutional Autoencoders
Dimensionality Reduction using Convolutional Autoencoders

A visualization in reduced dimension to show how autoencoders learn features in the Fashion MNIST dataset.

View Code
Fast Plane Extraction from a 3D Point Cloud
Fast Plane Extraction from a 3D Point Cloud

Fast, Parallel implementation of the RANSAC algorithm to segment a plane in a point cloud.

View Code
Post-Earthquake damage assesment using FCNs
Post-Earthquake damage assesment using FCNs

A convolutional neural network for multi-task visual damage assesment.

View Code
Two-View geometry based pose estimation
Two-View geometry based pose estimation

Relative Pose estimation using OpenCV

View Code
Power cable detection using Deep Learning
Power cable detection using Deep Learning

Convolutional Neural Network for detecting power cables.

View Code
Variable Computation in RNNs
Variable Computation in RNNs

A modification in a RNN unit that reduces redundant computation.

View Code
Tango+
Tango+

A tango inspired dark theme for Visual Studio Code.

View Code
Clean Portfolio
Clean Portfolio

A simple and elegant Jekyll theme for portfolio pages

View Code