Data Farming Intern

Role Overview

As a Data Farming Intern, you will play a key role in building and maintaining large-scale, high-quality datasets for robotics applications. You will work on data collection from real robots and simulations, annotation of multi-modal sensor data, data cleaning, and quality assurance directly contributing to improving robot perception, object detection, grasping, navigation, and manipulation capabilities.

This is an excellent opportunity for students passionate about robotics, computer vision, and AI data pipelines.

Key Responsibilities

• Collect and curate multi-modal robotic datasets including RGB images, depth maps, video sequences, LiDAR/point clouds, IMU, and joint telemetry from real robots and simulation environments (Gazebo, Isaac Sim, MuJoCo, etc.).

• Bounding boxes, segmentation masks, and keypoints for object detection and instance segmentation.

• 3D cuboids and semantic segmentation on point clouds.

• Action labelling, trajectory annotation, and temporal segmentation for manipulation and navigation.

• Pose estimation and tracking of objects/humans/robots.

• Clean, validate, and organize raw sensor data to ensure consistency, remove noise, and maintain high accuracy standards (target 95–98% QA accuracy).

• Support synthetic data generation and augmentation to increase dataset diversity and robustness (domain randomization, lighting variations, clutter simulation).

• Identify data gaps, biases, or edge cases (occlusions, low-light, dynamic scenes) and suggest improvements to the data farming pipeline.

• Collaborate with robotics engineers, perception scientists, and ML teams to refine data based on model performance feedback.

• Maintain proper documentation, version control of datasets, and metadata for traceability and reproducibility.

• Assist in running small data collection experiments with physical or simulated robots.Role & responsibilities

Preferred candidate profile