Intervention data is abundant, but review capacity decides what the next policy learns.
Fast curation protects iteration speed, while careful curation protects the policy from learning the wrong recovery.
Who you’d be doing this for
“I watched three hours of recoveries yesterday, and the same stall showed up again this morning.”
Maja Nielsen · Robotics Engineer
Runs manipulation cells and must decide which intervention episodes justify data curation before the next policy iteration.
What is at stake
Robotics engineers are reviewing hours of interventions, yet repeat failures still reach the next policy. You have to weigh faster data triage against evidence that the ranking selects genuinely useful failures.
Why it isn’t already fixed
Every obvious fix costs something else. That’s the part you’d have to decide.
- review speed vs. training-data relevance
- learned ranking vs. reviewer trust
- cross-robot reuse vs. setup-specific signals
- offline evidence vs. real-robot gains
Why Applied Intuition
Dana’s robot-learning workflow depends on turning real-robot interventions into reliable improvements across industrial tasks.
Written with these in mind
Not your kind of problem? 65 more at Applied Intuition, or browse every organization.
This is the setup. The work is inside.
Running it puts you in the room: the full situation and its constraints, stakeholders who push back in their own words, and the decisions that are yours to make. What you produce becomes a Day One Plan — work you can show someone instead of describing.