Robot demonstration review
Deepen AI's trained reviewers watch your robot demonstrations and sort them into useful and flawed, with the flaw noted, so your policy learns only from good examples.
Rate in your plan before you pay A plan within 2 working days No sales call
What a reviewed demonstration looks like
Each demonstration comes back useful, flawed or sent to you, with the moment that decided it.
- Demonstration
- Pick a mug from the rack and place it in the sink, episode 2204
- Operator marked
Success.- Reviewed as
- Flawed: regraspThe gripper slipped and regrasped the mug at 0:09 before placing it.
- Reason
- Your rules reject demonstrations with a regrasp for this task, so the policy doesn't learn the slip.
- Demonstration
- Fold a towel in thirds, episode 3150
- Operator marked
- Success.
- Reviewed as
- UsefulOne clean attempt, finished in 41 seconds.
- Reason
- Inside your time limit, with no pause longer than 3 seconds.
- Demonstration
- Open the top drawer, episode 3302
- Operator marked
Failure.- Reviewed as
- Useful: recoveryThe gripper missed the handle, then corrected and opened the drawer.
- Reason
- Your rules keep clean recoveries as their own category.
- Demonstration
- Hand a tool to a person, episode 3410
- Operator marked
- Success.
- Reviewed as
- Sent to youThe person took the tool before the gripper opened. Does a hand-over the person finished count as a success?
- Reason
- Your answer is added to the review rules.
How we start on your work
No sales call. A plan in your inbox, then a short paid pilot on your real demonstrations.
What we need from you
- Your demonstration recordings: video, joint states or both
- Your rules for a useful demonstration
- 20 to 30 demonstrations you've already reviewed
- 5 minutesTell us the task
What your robots do, how many demonstrations a week, and the categories you need.
- Within 2 working daysGet a plan and a price
Reviewers, hours, the rate, start date and the total price.
- 2 weeksRun the pilot
We train reviewers on your rules, review real demonstrations and send two weekly reports. Then you decide.
Who reviews, and how decisions stay consistent
Trained reviewers, working in English. Each one learns your rules and reviews a set of known demonstrations before live work.
Every reviewer's accuracy is measured
Items you've already answered are hidden in the real work, so we know how accurate each person is.
Nobody guesses on unclear cases
The team lead decides, or asks you. Your answer joins the guidelines.
You see the results every week
A weekly report shows work done, QA results and what's improving.
Standard review covers manipulation, mobile and humanoid demonstrations from video and logs. Review that needs robotics engineers is quoted in your plan.
How much does robot demonstration review cost?
Demonstration review is quoted in your plan, because the rate depends on how long your demonstrations are and the expertise they need. You pay for the hours reviewers work, with a team lead, QA checks and a weekly report included.
Set before you pay. Your plan shows the rate, team size, hours and total price.
Starts with a two-week pilot on your real demonstrations, paid upfront, with two weekly reports.
Your data stays protected. We work under NDA with the access you set. Deepen AI holds SOC 2 Type II, ISO 27001, ISO 9001 and TISAX.
Questions
What do reviewers look at?
Everything your recording has: video from each camera, joint states, gripper state and timing. Reviewers step through the moments that matter frame by frame.
Which categories can you use?
Yours. Common ones are useful, flawed with the type of flaw, clean recovery, and unsafe. Every decision notes the moment that decided it.
Can reviews help our operators improve?
Yes. The weekly report lists flaws by type and by operator, so the next collection sessions avoid them.
Can reviewers keep up with collection?
The team is sized to your weekly volume in the plan. Tell your team lead when collection ramps up, and reviewers are added.
What format are reviews delivered in?
Your schema, such as JSON lines with the episode ID, the category, the flaw type, the timestamp and the note.
How is our data protected?
We work under NDA, with the access you set, and you can remove it at any time. Deepen AI holds SOC 2 Type II, ISO 27001, ISO 9001 and TISAX.
Train only on demonstrations worth learning from
A plan within 2 working days. No sales call.