Abhinav Reddy Kowkuntla

Abhinav Reddy Kowkuntla

Perception | Navigation | Sensor Fusion | Robotics

I am a Graduate Research Assistant at the Northeastern Field Robotics Lab, passionate about SLAM, Computer Vision, and Deep Learning. I specialize in building robust perception systems for autonomous robots, from underwater mosaicing to multi-robot SLAM.

Core Expertise

Visual SLAM

  • Frontend & Backend Development
  • Bundle Adjustment & Factor Graphs
  • Multi-Map Generation
  • Loop Closure Detection

Deep Learning

  • Transformers & Attention Mechanisms
  • Vision Transformers (ViT)
  • Neural Rendering (NeRF, Gaussian Splatting)
  • Generative Models (GANs, Diffusion)
  • TensorRT Optimization

3D Perception

  • Stereo Camera Calibration
  • Depth Estimation & Disparity
  • Sensor Fusion (LiDAR + Camera)
  • Point Cloud Processing

Sensor Stack

  • LIDAR (Velodyne, Ouster)
  • IMU (MicroStrain, Xsens)
  • Cameras (Realsense, FLIR)
  • GPS (RTK, Differential)

Work Experience

Jan 2024 - Present

Northeastern Field Robotics Lab

Graduate Research Assistant | Boston, MA

  • Designing custom multimodal stereo setup (RGB + Thermal) with unified calibration.
  • Building end-to-end deep learning pipelines for dense correspondences and depth estimation.
  • Developed lightweight Transformer-based alternative to RAFT-Stereo (14x faster inference on Jetson Orin Nano).
  • Supporting calibration and ROS bag pipelines for multi-sensor data.
Jan 2021 - Dec 2023

Robert Bosch

Function Owner / Lead System Developer | India, Hungary

  • Led development of ADAS software (Automatic Parking) in Simulink/C++ for Honda, Nissan, Mazda.
  • Delivered $2M+ in production features compliant with AUTOSAR.
  • Improved efficiency by automating workflows, reducing release timelines by 1-2 weeks.
  • Supervised GitHub/Jenkins for a team of 20 associates.

Featured Projects

Multi-Robot SLAM

Multi-Robot SLAM & Dataset

Large-Scale Data Collection & Benchmarking

Cross-Modal Sync

Cross-Modal Synchronization

Infrared–RGB Stereo Depth Transfer

Underwater Mosaicing

Large-Scale Underwater Mosaicing

GTSAM factor graph optimization

3D Reconstruction

Multi-view 3D Reconstruction

Bundle Adjustment & Sparse Reconstruction

Underwater Segmentation

Underwater Image Segmentation

Self-Supervised Learning with SimCLR

Depth Estimation

Efficient Depth Estimation

RAFT-Stereo & NEUFlow-V2

All Projects

Cross-Modal Sync

Cross-Modal Synchronization

Infrared–RGB Stereo Depth Transfer using Teacher-Student distillation.

Calibration Stereo Deep Learning
Multi-Robot SLAM

Multi-Robot SLAM & Dataset

Deployed fleet of 5 robots with RGB/LiDAR/IMU for large-scale data collection. Synchronized 1TB+ data for SLAM benchmarking.

ROS2 SLAM Robotics
GPS-IMU Fusion

GPS-IMU Fusion

Implemented 2D pose estimation pipeline fusing GPS and IMU data using Extended Kalman Filter (EKF).

Python Kalman Filter Sensor Fusion
Mosaicing

Underwater Mosaicing

Homography & GTSAM factor graph optimization for 200+ images.

OpenCV GTSAM
SFM

3D Reconstruction

Sparse 3D Reconstruction using Bundle Adjustment.

Python OpenCV
Segmentation

Underwater Segmentation

Self-Supervised Pre-trained model using SimCLR & U-Net.

PyTorch Deep Learning
Depth

Depth Estimation

RAFT-Stereo and NEUFlow-V2 integration.

DL Stereo
RCM

RCM Healthcare Robot

Remote Center of Motion robot model with kinematic analysis.

SolidWorks
SLAM

SLAM Analysis

Comparative Analysis of Lio-Sam and Lego-Loam.

ROS Gazebo