pk.

Robotics & machine intelligence

Pranav
Kumar.

Intelligence, out in the real world.

I’m a Computer Science student at Purdue, working on machine learning, perception, and robot control. I build systems that turn what robots see into what they do.

Explore my research GitHubLinkedIn
Currently atSCALE Robotics LabVibe Robotics
Pranav with a humanoid robot at an AMD robotics demonstration
Somewhere between research & reality.FIG. 01
A little newsOur VLA interpretability paper was accepted to NeurIPS.Read the paper

01 / Research & current work

From understanding
to action.

How do robots perceive their surroundings, learn useful behaviors, and act reliably beyond the lab? That’s the thread through my work.

NeurIPS — Neural Information Processing SystemsResearch publication
Accepted at NeurIPSVLA · Interpretability

Mechanistic Interpretability and Steering of VLA Models Through Sparse Autoencoders

Understanding what Vision-Language-Action models learn, and how we can steer their behavior. Co-authored research on mechanistic interpretability using sparse autoencoders.

Read the paper on arXiv
Vibe Robotics01

Real-time perception and hybrid control for humanoid robots

Real-time RGB-D perception, spatial mapping, and hybrid control that brings teleoperation and autonomous policies together.

RGB-D perceptionSLAMHybrid control
More about this work
  • Robotics Engineer Intern — RGB-D perception, point clouds, and SLAM-based spatial modeling for humanoid interaction.
  • Build real-time perception pipelines that turn raw sensor streams into structured scene representations.
  • Develop hybrid control systems combining teleoperation with autonomous policies for closed-loop execution.
Vibe RoboticsVibe A1
SCALE Robotics Lab02

Vision-Language-Action models for closed-loop robot control

Learning robot policies from multimodal inputs, studying their failure modes, and connecting perception to closed-loop action.

VLA modelsImitation learningSim-to-real
More about this work
  • Trained VLA models (Open-Pi, OpenVLA, ACT) for closed-loop robotic control from multimodal inputs — imitation learning and RL for policy refinement.
  • Built perception-to-action pipelines integrating RGB-D sensing, SLAM, and state feedback for real-time inference.
  • Conducted sim-to-real experiments analyzing generalization gaps, failure modes, and robustness of learned policies.
  • Built a real-world Connect 4 robotic system combining VLA policies with classical control (Purdue Robotics Day demo).
  • Co-authored research on mechanistic interpretability and steering of VLA models through sparse autoencoders (SAEs).
  • Working on hallucination mitigation in video models for World Action Models.
SCALE Robotics LabVLA paper

02 / In the field

Real robots.
Real learning.

The interesting part happens when a model meets the physical world. A few moments from building, testing, and figuring things out.

Pranav demonstrating robot manipulation and Connect 4 at Purdue Robotics Day

Field notes / Purdue Robotics Day

Taking research
off the bench.

A real-world Connect 4 system that brings learned VLA policies and classical control together. There’s nothing quite like a live demo to find out what actually works.

Watch the robots in action Perception → Policy → Action
Vibe Robotics01

Humanoids, hands-on

Building perception and hybrid control for humanoids. Built and trained in house — literally.

2× speed
Policy data collection02

Your move, SPOT.

A Connect 4 opponent, and a real-world data collection pipeline for policy training.

Vision-language-action03

A policy at play

Pi0.5 takes on Connect 4: learned policies, game interaction, and embodied decision-making.

A closer look / Selected projects

01
robot perception

RealSense to Unity 3D Perception & SLAM Pipeline

I built a real-time RGB-D pipeline from an Intel RealSense D435i for 3D reconstruction, SLAM-based localization, and spatial mapping in Unity.

02
VLA + classical control

Connect 4 Robot Demo

I built a real-world Connect 4 robotic system that combined learned VLA policies with classical control for Purdue Robotics Day.

03 / Experience

Different problems. Same curiosity.

From scientific ML to social systems and aerial robotics, each experience shapes how I build.

01

Dow

Agentic ML workflows for scientific analysis

Build agentic ML workflows with LangGraph and DSPy for automated scientific analysis, instead of one-off prompts. Focus on reproducibility — traceable, repeatable model outputs that scale beyond a single analysis run.

LangGraph, DSPy, scientific ML pipelines
02

Contextual Intelligence

Event pipelines and graph-based interaction modeling

Software Engineer Intern — re-architected high-throughput event ingestion pipelines for product behavior data. Designed graph-based interaction models for large-scale prediction systems, bridging product engineering and ML.

event ingestion, graph models, prediction systems
03

Purdue Aerial Robotics Team

UAV perception and control systems

Worked on sensor fusion pipelines for stable real-time UAV flight. Early robotics systems work before moving deeper into ML for robot control.

UAVs, sensor fusion, real-time flight

04 / The toolkit

Tools for the job.

From training a policy to getting a robot to execute it.

Languages

Python · C/C++ · Java · R · JavaScript · TypeScript · SQL

Robotics / AI

ROS · SLAM · RGB-D Perception · Point Clouds · Sensor Fusion · VLA · WAMs · Isaac · MuJoCo

ML

PyTorch · TensorFlow · Reinforcement Learning · Imitation Learning · Computer Vision · LangGraph · DSPy

Systems

Docker · Linux · Flask · Node.js · React · MongoDB · Unity

Along the way / Recognition

A few highlights.

HackMIT

Citadel Challenge Winner

2025

NASA SUITS

Best Innovative

2025

Catapult Hackathon

Runner-up

2025

HackMIT

Top 12 / 250+

2024

Boilermake XII

Top 8 / 109

CS Base Climate Hackathon

Advance Tier Second Place

2024

Purdue Hello World

2nd place

2023

Indian National Mathematical Olympiad

National Team

Keep in touch

Good things start
with a conversation.

Robotics, research, or something worth building.
I’d love to hear what you’re working on.

cdtpranavkumar1126@gmail.com