Robotics system engineer

Saakh - AI-first B2B intelligence platform; Ahmedabad District; On-site; INR 558,000 - 837,000; Full time

Job summary

Saakh in Ahmedabad is seeking a Robotics Systems Engineer to build a live robotics system on Jetson, LiDAR, cameras, and sensors.

You'll own perception and autonomy stack, integrate ROS 2, verify reliability, and ship a real robot in a hands-on role that values proven builds.

Join a practical, fast-paced lab where six days on-site and immediate hands-on work are expected. Education is useful but concrete builds count more.

Qualifications

• Non-negotiable Python and Linux proficiency.
• Experience with LiDAR, depth cameras or RealSense.
• Demonstrated ability to deploy models on constrained hardware.

Responsibilities

• Own the intelligence layer for perception, navigation and response.
• Move models from laptop to robot using Jetson and ROS 2.
• Build ROS 2 architecture and reusable components.
• Ensure boot, deployment, telemetry, and robust operation.

Skills

Perception

Autonomy

Integration

ROS 2

Python

Linux

Git/GitHub

LiDAR

RealSense

MQTT

Tools

Jetson

Docker

Isaac Sim

GStreamer

ROS 2

CAN bus

STM32

Job description

Ref: MYSA/HR/2026-27/ROB-111 Date: 3 October 2026

Dear Future Robot Builder,

We are opening a Robotics & Physical AI Lab in Ahmedabad, and right now it is a table. On that table are NVIDIA

Jetson boards, cameras, a LiDAR, motors, microcontrollers, a Bambu Lab H2D AMS Combo and a DGX Spark in the

corner. What it does not have yet is a robot.

That is the job. Two of you, turning that table into machines that can see, navigate, speak, and keep working after you have

gone home. Mysa builds AI-first products from Ahmedabad - Saakh, India's AI-led B2B intelligence platform, Swar for

voice-first medical records, MamaMate for the Indian household. All of them live on a screen. This one will not.

We care far more about what you have built than how long you have worked. Six months and an autonomous Jetson robot

on video beats four years of generic application development. That is not a platitude - it is literally how we will sort the

applications.

The bits you'll read twice.

Position Robotics Systems Engineer · Robotics & Physical AI Lab

Seats 2 open. Apply once; we will place you.

Two ways in (a) Full-time, 1-3 years of relevant building. (b) Internship-to-full-time, six months paid, for

exceptional freshers and self-taught builders, converting on performance.

Salary (CTC) Full-time: Rs. 50,000 - 75,000 per month, placed on evidence, not on years.

Internship lane: Rs. 25,000 per month for six months, then the band above.

Reports to Arpit Golechha, Founder - directly. No layer between you and the decision.

Location Ahmedabad · on-site · six days a week. Robots do not debug over video call.

Equipment All provided - Jetson devices, cameras, LiDAR, sensors and microcontrollers, a Bambu Lab H2D

AMS Combo on the bench, and access to a DGX Spark and GPU workstations for training and

simulation.

Education Useful, not mandatory - CS, Robotics, Electronics, Electrical, Mechatronics, AI, Mechanical or

Embedded Systems. We will judge the robot, not the certificate.

The two seats.

Both sit in the same stack and both are hands-on. We split them so two people do not solve the same problem twice. In

practice the line will blur.

Seat 1 Seat 2

Perception & Autonomy - make the robot see and decide.

Camera and LiDAR pipelines, person and object detection,

OpenCV, YOLO, Isaac ROS.

Integration & Reliability - make the robot work.

Jetson-to-microcontroller interfaces, motor and encoder

control, boot and deployment, video streaming, telemetry,

STM32, UART / I2C / CAN, GStreamer, MQTT.

What you'll actually do.

• Own the intelligence layer. Someone else may bolt the chassis together and wire the motors. Making the machine perceive, navigate, decide and respond is yours.
• Get models off the laptop and onto the robot. PyTorch to ONNX to TensorRT to Jetson to ROS 2 - and the hard part is the last three steps: latency, GPU memory, FP16 and INT8, thermals, frames you cannot afford to drop.
• Build the ROS 2 architecture we will reuse. Nodes, packages, launch files, services, actions, TF2, URDF, Nav2. A platform, not fifteen one-off robots.
• Make it boot. A robot that needs an engineer to open five terminal windows is a university demonstration. Docker,
• Simulate before you break things. Isaac Sim and Isaac Lab, so the first real‑world test is not the first test.
• Diagnose the physical world. USB disconnects, motor noise resetting a microcontroller, localisation drift, thermal throttling, a node dying when Wi‑Fi drops. This is most of the job on most days, and it is the part we are really hiring for.

What we need you to have already.

Non-negotiable Python. Linux you are genuinely comfortable in, not just a terminal you have seen. Git and GitHub.

Enough networking to debug a connection yourself. The habit of reading a log before asking a

question.

LiDAR · stereo or depth cameras · embedded systems. We do not expect all of them - we expect

real use of some.

RealSense, ESP32, STM32, FreeRTOS, CAN bus, MQTT, visual or LiDAR SLAM, Kalman

filters, mobile robots, robot arms, drones, reinforcement learning, vision-language-action models.

What counts as proof A repository we can read. A video of the thing moving. A write-up of what broke. College team,

hackathon, personal build, internship - the origin does not matter.

The questions you should enjoy.

• Why is the Jetson dropping camera frames?
• Why does TensorRT inference get slower after ten minutes?
• Why is the robot oscillating while following a person?
• Why is Nav2 refusing to go through a doorway it can clearly fit through?
• Why does motor noise reset the microcontroller?
• Why did a ROS node die when Wi‑Fi dropped, and why did nothing restart it?
• Why is the robot seeing itself as an obstacle?

This is probably not for you if.

• Your AI experience is calling other people's APIs.
• You have trained models but never deployed one onto constrained hardware.
• Your robotics is coursework and simulation with no physical build behind it.
• Your embedded experience stops at following Arduino tutorials.
• You are a mechanical CAD or PCB specialist - both are valuable here, neither is this role.
• You want to work remote. We have tried to imagine debugging LiDAR drift over a call and we cannot.

We print our own parts.

There is a Bambu Lab H2D AMS Combo in the lab and it is not decoration. We are unreasonably fond of 3D printing

Here. When a sensor needs a mount, a LiDAR needs a riser, a camera needs a bracket at one particular angle or a battery

needs a tray, you design it in the afternoon and you are testing it on the robot the same evening. No vendor, no three‑week

lead time, no waiting in someone else's CAD queue.

This matters more than it sounds. Most robotics iteration is throttled by how long a physical part takes to arrive. Ours is

throttled by how fast you can think. If you already design your own brackets you will feel at home on day one; if you have

never touched CAD, you will learn, and it will change how fast you work permanently.

How to apply.

Write to [e-mail] with the subject line:

Robotics Systems Engineer - [Your Name] - [Seat 1 / Seat 2 / Either]

Five things in the body. Keep it short - we would rather read a tight email and a long repository than the reverse.

1. Your resume, attached.

2. Your GitHub, and which repository we should actually open.

3. A link to a video of something you built, moving. Phone footage is perfect. A breadboard mess is perfect. We care

that it worked.

4. One paragraph on the hardest bug you have personally fixed in a physical system - what you thought it was, what it

turned out to be, and how you found out.

5. Which seat, which lane, and whether you can be in Ahmedabad.

What happens next.

A short call for the basics, then a 45-minute build round with the founder - not a whiteboard, not a

puzzle round, but your actual build and the decisions inside it. Shortlisted candidates get the full role specification first. A

decision follows within 48 hours of the call.

If you have built something real and nobody has looked at it properly yet, this is an invitation to be taken seriously.