Case Study · Robotics Data Collection
Running a Teleoperation Data Floor for a Frontier AI Lab
A leading AI research lab needed vast volumes of high-quality robot demonstration data — captured on their own platforms, at their own facilities, at production scale. FS Studio built and now runs the entire teleoperation operation as a managed service, so the lab’s engineers stay on the models instead of the floor.
The Challenge
The models were ready. The data pipeline wasn’t.
Training robots to perform real-world tasks reliably takes enormous volumes of structured teleoperation data — captured by skilled operators driving real robots, shift after shift, to a consistent standard.
The lab had the platforms, the facilities, and the AI programs. What it didn’t have was a way to run a data-collection floor at scale without pulling its own engineers into staffing, scheduling, certification, coverage, and daily reporting. Standing that operation up in-house — and keeping it staffed every shift, with trained operators who meet a strict competency bar — is a full operational discipline of its own.
They needed a partner who could own that discipline end to end: deploy the people, hit the throughput and quality targets, and integrate cleanly into the lab’s own workforce and systems — without the lab ever having to manage the floor.
The Solution
A fully managed teleoperation operation — FS Studio owns the floor.
FS Studio deploys trained teleoperators to run the client’s robot platforms on-site, generating structured training data to the lab’s spec — and takes ownership of everything it takes to keep that floor running to standard.
Two-person cell model
Each active robot is staffed by a primary operator running teleoperation tasks and an observer/reset technician managing fault response, sensor monitoring, and session logging. It’s the operational unit the whole program scales on.
Tiered operator bench
A layered talent bench — from operational teleoperators to robotics-aware seniors, advanced technicians, and on-site team leads — so complex fault scenarios and escalations are handled without slowing the line.
SOP training & certification
No operator touches an active robot before completing the client’s required SOP training and meeting competency standards. FS Studio maintains the training records and re-certifies operators within 48 hours whenever SOPs change.
Reporting & incident cadence
Daily shift reports, weekly performance summaries, and rapid incident reporting through the client’s tracking dashboard — backed by repair tickets, state-change logs, and workstation health logs.
Continuous shift coverage
FS Studio guarantees continuous operation of every assigned robot through the shift — absorbing coverage for breaks, absences, and meals so the client sees zero unstaffed robot time.
Integrated workforce
Operators work alongside the client’s own staff in a shared environment — following site rules, coordinating robot assignments and shift handoffs, and touching only the systems their tasks require.
How We Work
A shared-responsibility model, tightly defined.
The engagement works because the line between client and studio is drawn clearly. The client owns the platforms, SOPs, and data infrastructure; FS Studio owns the people and the operation. Each side’s obligations are explicit — which is exactly what lets the floor run predictably.
Stand up the operation
FS Studio sources, screens, and onboards the operator bench, then integrates into the client’s facility — badging, safety orientation, and access — ready to run assigned robots from day one of each phase.
Certify against the client’s SOPs
Every operator completes client-required SOP training and meets the competency bar before going live. Training records are maintained and produced on request; re-training follows any SOP revision within 48 hours.
Run the floor to standard
Two-person cells run continuous shifts against clear performance goals — raw hours per shift, zero unstaffed robots, and successful-episode quality — with a team lead owning coverage, attendance, and daily logs.
Report, escalate, improve
Daily and weekly reporting keeps the client fully visible on throughput, quality, staffing, and risk. An engagement manager owns performance reporting and escalation — and underperformers are replaced fast when flagged.
Built to Scale
Pilot to program, phase by phase.
The model is designed to prove out small and scale on demand. Each phase gates the next: the client authorizes expansion only once the operation has demonstrated it can hit targets — and the same blueprint extends to new tasks, new facilities, and new robot counts.
Pilot & ramp
A focused deployment stands up the first work cells, proves the SOP and reporting loop, and establishes the quality baseline on a small fleet of robots.
Prove the model
Authorized expansion
On the client’s go-ahead, the operator bench and robot count scale up materially — more cells, more shifts, same standard — under a single managed program.
Scale the floor
New tasks & sites
The blueprint travels. Additional task types, facilities, or parallel engagements spin up on the same operating model, coordinated through change orders.
Replicate anywhere
What “Good” Looks Like
Operations run against explicit targets.
Performance isn’t a vibe — it’s measured every shift against goals the client and FS Studio agree up front, so quality and throughput stay visible and accountable throughout the engagement.
The client owns the science and the robots. FS Studio owns the people, the shifts, and the standard — so the data keeps flowing and the research team never has to run a floor.
— The FS Studio operating model
The Outcome
Clean data at scale — and a floor the client never has to run.
FS Studio turned a daunting operational build into a running, measured, scalable service — and the engagement grew from an initial pilot into an ongoing multi-phase program.
Structured data, to specTeleoperation session data captured in the client’s format, every shift.
Zero floor overheadThe lab’s engineers stayed on the models, not staffing and logistics.
Scaled on demandRobot count and operator bench grew phase over phase, on the client’s cue.
A repeatable blueprintThe same model now extends to new tasks, sites, and engagements.
Need a teleoperation data floor stood up and run?
FS Studio builds and operates robotics data-collection programs end to end — trained operators, certification, QA, and reporting, scaled to your platforms and your standard.