Hi, I am a Master's student in Mechanical Engineering (Robotics and Controls) at the University of Colorado Boulder. I completed my Bachelor's degree in Mechatronics Engineering with a strong focus on robotics.

My expertise and interests include:

★ -- Primary Focus
Mechatronics★ Computer Vision State Estimation Navigation & Planning★ Robot Learning★ MLOps/DevOps★

My focus is on designing & building intelligent robots, implementing Robot learning techniques to enhance their autonomy, and deploying scalable robotic systems to solve real-world problems. Aspiring to be a Robotics Generalist.

In my spare time, I enjoy exploring astronomy, stargazing and working out at the gym.

I am open to work in:

Robotics Software Engineer Robot Learning & Autonomy Roles MLOps/DevOps Engineer
View Resume Download Resume
Santosh Chaganti

Experience

Teaching

Course Assistant

University of Colorado Boulder

January 2025 - May 2025

Served as Course Assistant for MCEN 1030 - Introduction to Engineering Computing. Supported students in learning programming fundamentals(MATLAB & Python) and engineering applications.

Teaching Engineering Computing Education Programming
Teaching

Grader

University of Colorado Boulder

January 2025 - May 2025

Graded assignments and provided feedback for MCEN 6228 graduate-level engineering course.

Grading Graduate Course Academic Assessment
Research

Team Member

University of Delaware (Remote + Fieldwork)

June 2024 - August 2024

Deployed Slocum Glider hardware in ocean environments for data collection. Led development of a Particle Filter in Python for underwater localization. Designed and validated system model pipeline including data cleaning and performance evaluation.

UnderwaterRobotics ParticleFilter Localization Python DataAnalysis
Research

Research Assistant

University of Colorado Boulder, HIRO Lab

January 2024 - May 2024

Developed motion planning algorithms for Franka Emika robot using ROS and C++. Co-authored sections of a paper accepted to IROS 2024. Visualized experimental data using R and RViz to support rapid iteration.

MotionPlanning ROS C++ FrankaEmika IROS2024
Leadership

Chapter Secretary

IEEE Denver Computer, Information Theory, and Robotics Society

October 2023 - October 2024

Served as the official representative for IEEE CIR Society at high-level Regional Committee (EXCOM) meetings, conveying chapter priorities and reporting outcomes. Orchestrated monthly technical presentations featuring industry experts and coordinated internal organizational meetings. Maintained meticulous documentation of meeting proceedings, action items, and decisions for both internal operations and external partnerships

IEEE Leadership Technical Presentations Documentation Community Engagement
Leadership

Board Member

Engineering Excellence Fund

October 2023 - August 2024

Served on the board for the Engineering Excellence Fund, participating in decision-making processes for resource allocation and fund management.

Board Member Fund Management Engineering Resource Allocation
Industry

Senior Analyst, Data Engineering

LatentView Analytics, India

August 2021 - July 2023

Led Apache Airflow migration from Astronomer to MWAA, optimizing 100+ workflows and saving $200K annually. Developed Python-based Airflow pipelines and SQL-based DBT models with GitHub Actions CI/CD automation. Built data governance framework standardizing lineage and access controls. Set up SageMaker infrastructure for ML model inference pipelines and created Grafana dashboards for AWS resource monitoring.

DataEngineering AWS Airflow Python MLOps DBT GitHub Actions SageMaker Grafana

Projects

Course Project

Cooperative Air-Ground Robot Localization

GitHub Report

Developed a cooperative localization system for a UGV-UAV team using Linearized (LKF) and Extended Kalman Filters (EKF). Implemented state estimation with 95% NEES consistency, achieving <2.5cm position error for ground vehicles and <5cm for aerial vehicles. Derived Jacobians for nonlinear dynamics, tuned noise parameters (Q/R scaling), and validated performance via Monte Carlo simulations in MATLAB.

State Estimation Kalman Filtering Sensor Fusion Cooperative Localization LKF EKF Monte Carlo Simulations MATLAB ODE45
Course Project

Multi-View Generation from Single Image Using 3D Reconstruction and Inpainting

Colab Report

Developed a pipeline to synthesize novel viewpoints from a single stereo image using RAFT-Stereo for depth estimation, 3D reconstruction, and LaMa inpainting for occlusion handling. Generated realistic multi-perspective renders by transforming camera poses and filling occluded regions, demonstrating applications for VR/AR content creation.

3D Reconstruction Depth Estimation Disparity Mapping Image Inpainting RAFT-Stereo LaMa OpenCV Python
Course Project

Hybrid Systems Final Project

GitHub

Implemented A* and Q-learning algorithms for LTL-constrained pathfinding in a grid world. Demonstrated A*'s efficiency for deterministic planning and Q-learning's adaptability for dynamic environments, highlighting trade-offs between reliability and flexibility for autonomous navigation.

A* Q-learning LTL Formal Verification Reinforcement Learning Path Planning
Course Project

Autonomous Dueling Tank Robot

Report

Designed an autonomous combat robot with mecanum wheels, computer vision targeting (OpenCV/PixyCam), and a flywheel launcher. Integrated Teensy 4.1 and Raspberry Pi on a custom PCB for motor control and decision-making, overcoming PCB overheating challenges.

C++ Python PID Control Inverse Kinematics OpenCV PCB Design Motor Control Sensor Fusion Teensy Raspberry Pi
Course Project

Deep Reinforcement Learning for Autonomous Racing Control

GitHub Report

Developed a Deep Q-Network (DQN) agent for autonomous racing in OpenAI Gym's CarRacing environment. Implemented CNN-based state processing (96x96 RGB images) and discrete action control (steering/acceleration/braking), achieving stable performance (600–800 reward) through hyperparameter tuning and experience replay. Compared different architectures to optimize training stability and computational efficiency.

DQN Q-Learning PyTorch Robotic Control Hyperparameter Tuning Deep Reinforcement Learning Discrete Action Spaces Reward Shaping
Course Project

Mobile Manipulation using Motion Planning

GitHub Report

Implemented motion planning for a mobile manipulator (ROS/Gazebo) to autonomously navigate to target tables (95% success rate) and perform pick-and-place tasks (90% grasp accuracy). Leveraged ArUco markers for precise localization (2.5 cm position error) and MoveIt! for collision-free arm trajectories. Demonstrated modular integration of perception, navigation, and manipulation, laying groundwork for LTL-based task sequencing.

Mobile Manipulation Motion Planning Collision Avoidance ROS Gazebo Localization Inverse Kinematics MoveIt! OMPL
Implementation

Motion planning Techniques Coursework

Implemented various motion planning techniques as part of my coursework at University of Colorado Boulder.

Mobile Manipulation Motion Planning Collision Avoidance CPP Localization Inverse Kinematics MoveIt! OMPL
Implementation

Coded Deep Depth

Colab Part 1 Colab Part 2 Report

Implemented a U-Net-based pipeline for monocular depth estimation using coded RGB images synthesized via depth-dependent PSF convolution. Achieved Rel-Abs Error: 0.0575 and RMSE: 0.0405 on NYUv2 and UMD datasets, with real-time inference at 212 FPS. Implemented depth-aware blur simulation (Phase I) and trained a frequency-loss-optimized U-Net (Phase II), demonstrating robustness across in-domain and out-of-domain scenes.

Monocular Depth Estimation Coded Imaging PyTorch PSF Convolution U-Net Deep Learning Computer Vision
Implementation

Blob The Builder - Gaussian Splatting

GitHub Report

Implemented a Structure-from-Motion (SfM) pipeline to reconstruct 3D scenes from 2D images using geometric computer vision techniques. Achieved accurate 3D point cloud reconstruction via robust feature matching (RANSAC), fundamental/essential matrix estimation, and bundle adjustment. Demonstrated pose estimation (PnP) with cheirality constraints and non-linear triangulation, validated on multi-view street scene imagery.

SfM 8-Point Algorithm SVD 3D Reconstruction Epipolar Geometry Bundle Adjustment PnP Computer Vision
Course Project

NLP Implementations

GitHub

Implemented various NLP algorithms and models as part of my coursework at University of Colorado Boulder. This repository contains implementations of popular NLP models and techniques, including:

NLP PyTorch Transformers Sequence Models Attention Mechanisms Language Models
Implementation

LeRObot Implementation

Build SO 100 ARM robot and implemented a lerobot ACT and robot learning models

Python Foundation Models Robot Learning
Research Project

Autonomous Systems Bootcamp 2024

Developed predictive models for underwater glider trajectory estimation using steady-state equations and kinematic frameworks. Implemented physics-based algorithms incorporating critical parameters such as mass distribution and buoyancy dynamics to predict glider positioning in GPS-denied underwater environments. Validated model accuracy through simulation and testing during the 2024 Autonomous System Bootcamp at the University of Delaware, providing researchers with reliable location estimation capabilities for deployed gliders.

Trajectory Prediction Underwater Navigation Position Estimation Kinematic Modeling Parameter Optimization Physics-Based Simulation Python Particle Filter

Publications

Exploring How Non-Prehensile Manipulation Expands Capability in Robots Experiencing Multi-Joint Failure

Authors: Gilberto Briscoe-Martinez, Anuj Pasricha, Ava Abderezaei, Santosh Chaganti, Sarath Chandra Vajrala, Sri Kanth Popuri, and Alessandro Roncone

Published in: IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2024)

This work explores non-prehensile manipulation and whole-body interaction as strategies for enabling robotic manipulators to conduct manipulation tasks despite experiencing locked multi-joint failures.

Read Paper