Robotics + AI Research InternLocation: Kanpur, Uttar Pradesh | In-person

Duration: 2 months

Stipend: ₹25,000/month

Start: As soon as possible

Apply here: https://forms.gle/PzYPRj8dskyAVBS66

About the workI’m Bharat Jain, and I work at the intersection of robotics, LLMs, AI agents, and embodied/Physical AI. I build and experiment with systems that allow modern AI models to interact with and control real robots, and regularly share my projects and work in Physical AI on my YouTube channel.

Right now, I’m working on open research around a specific question:

How effectively can up-to-date LLMs/VLMs understand, reason about, and control robotic manipulation tasks?

I’m looking for one intern in Kanpur to work closely with me on this for the next two months.

The primary experimental platform will be the SO-101 robotic arm, along with cameras, ROS/robotics tooling, AI models, and related hardware.

What you’ll work onYou’ll help design and run experiments involving robotic manipulation, including:

- Setting up and experimenting with the SO-101 robotic arm
- Testing different approaches for controlling robot manipulators using LLMs and VLMs
- Experimenting with new frontier models such as GPT-6 Astra and other vision-language models
- Comparing different architectures and approaches under controlled tasks
- Designing benchmarks and evaluation metrics
- Running repeated experiments and collecting structured results
- Debugging robotics hardware, software, cameras, APIs, ROS components, model integrations, etc.
- Writing Python code and small tools required for experiments
- Analysing failures and figuring out why a particular approach did or did not work
- Documenting experiments so they can be reproduced
- Helping build technical demonstrations of the research

If the experiments produce sufficiently interesting and rigorous results, we may work toward a research paper or technical report.



Research authorship would depend on the actual contribution made. Regardless, the goal is to produce technically strong experiments, benchmarks, open-source work, and demonstrations that can be shared publicly through my existing robotics/AI media channels.

Who I’m looking forThis role is less about having every qualification already and more about being someone who can figure things out.

You should ideally

- Be pursuing or recently have pursued Computer Science, AI/ML, Robotics, Electronics, Mechatronics, or a related field
- Be comfortable programming in Python
- Understand basic machine learning / deep learning concepts
- Have experience working with robotics, embedded systems, computer vision, ROS, or hardware-software systems
- Be comfortable debugging things when documentation is incomplete
- Be able to read GitHub repositories, papers, documentation, and unfamiliar codebases
- Be able to independently learn a new technology and turn it into something working
- Be comfortable physically working with robot hardware

Strong preference if you have worked with:
- SO-100 / SO-101 / LeRobot arms
- ROS / ROS 2
- Robot manipulators
- MoveIt / inverse kinematics
- Cameras and computer vision
- Vision Language Models
- LLM APIs / AI agents
- PyTorch
- Hugging Face / LeRobot
- Jetson devices
- Arduino / Raspberry Pi / embedded hardware

You do not need experience with all of these. What matters much more is evidence that you have previously built difficult technical projects and can learn unfamiliar things quickly.

What this internship is NOTThis is not primarily:

- A course or training program
- A frontend/web-development internship
- A role where tasks will always have step-by-step instructions
- A purely theoretical ML internship

A large part of robotics research is: “This should work, but it doesn’t. Figure out why.”

If that sounds exciting rather than frustrating, you’ll probably enjoy the role.

📌 Physical AI Research Intern (Kanpur)
🏢 Dazzle Orc
📍 Kanpur