Rajdeep Singh

I am a second year M.S. Robotics student at the GRASP Laboratory, University of Pennsylvania. I enjoy working on modern robot learning and ML driven systems, and believe equally strongly in building on a rigorous foundation of mathematics and classical robotics, spanning dynamics, estimation, and control. I see the two as complementary rather than competing approaches.

I consider myself a full-stack roboticist, and my 5+ years of professional software engineering background enables me to approach SOTA robotics research problems with a strong focus on engineering robust, production-quality systems.

My current research focuses on Knowledge Based Neural ODEs for predictive control and residual dynamics learning. This summer (2026), I am working with Prof. M. Ani Hsieh on learning residual dynamics for underwater manipulators. Previously, I worked with Prof. Antonio Loquercio on DARPA's Abstract Planning and Semantic Reasoning (APSU) Challenge, developing autonomous robot navigation and semantic reasoning agents with Sim2Sim transfer capabilities.

In my past life, from 2024 to 2025, I was a Research Staff member at the DREAM Lab at the Indian Institute of Science, advised by Prof. Yogesh Simmhan, building cloud-edge hybrid systems for ML driven robotics applications. Prior to that, I spent five years at (2018-2023) Loco as an SDE-3, leading the data engineering and machine learning infrastructure powering production systems serving millions of daily active users.

Email  /  GitHub  /  Google Scholar  /  LinkedIn  /  Medium  /  Youtube

Research & Publications

AeroDaaS: Towards an Application Programming Framework for Drones-as-a-Service

We show how shape reconstruction with 3D Gaussians can be expanded to include differentiable optical flow, colored mesh exports and more.

Medium Publication: Robotics Demystified

I love to write, building a technical publication focussed on Robotics.

Projects

These include coursework, side projects and unpublished research work.

Poseidon: Underwater Manipulator Arm

A 3 DoF arm for which I wrote the software stack all the way from Dynamixel motors to high level.

Franka Emika Panda: Vision based Pick-and-Place

On a Franka Emika Panda hardware. With Chris Atkeson and Alex Spitzer. Using optimizers to match an observed trajectory.

Classifying Robotic Grasps

With Chris Atkeson and Alex Spitzer. Using optimizers to match an observed trajectory.

Design and source code from Jon Barron's website