Job Description

Role Overview : You will be taking raw footage of manual workers and turning it into clean, structured, labelled data that robotics companies can use to train their robots. Quality is everything in this role — bad data is worse than no data. 

Key Responsibilities: 

• Research what training data formats and annotation standards robotics companies actually need
• Set up and manage a video labelling workflow using standard tools (e.g. Label Studio, CVAT, Roboflow)
• Label raw footage clips — tagging body keypoints, action types, and work context
• Build and run a QC checklist on each data batch before it goes to clients
• Flag and manage rejected footage — document reasons and feed back to the hardware team
• Package labelled datasets in the format requested by each robotics partner
• Document the full labelling process so it can be handed to gig workers at scale
• Present findings at Month 2 and Month 4 milestone reviews

Job Requirements

Qualifications: 

• Studying or self-taught in data science, AI/ML, computer vision, or a related field
• Has used at least one data labelling or ML tool before (anything counts)
• Detail-oriented — you naturally spot errors others miss
• Comfortable doing research independently and synthesising what you find
• Uses AI tools naturally as part of how you work

Additional Advantages:

• Familiar with annotation frameworks such as COCO or OpenPose
• Has done any kind of data cleaning, structuring, or pipeline work before
• Knows basic Python — enough to run or tweak a script

Skills