Company
We sell capture programs, not a data lake.
Validated hours delivered
how we count ›
Hours that passed sync + QA and shipped to a buyer. Recaptured and rejected hours excluded.
Median cross-stream sync error
how we count ›
Median absolute timestamp offset between head camera and every other stream, over 12,480 measured frames.
Documented consent
how we count ›
Every participant signs a written commercial release before capture. Bystanders are de-identified or the clip is discarded.
Median time to first delivery
how we count ›
Signed spec to first validated batch in your hands. Measured across Spec Pilot engagements in environments where coverage is already deep — thin scenarios take longer to staff.
How we operate
Five commitments that cost us money.
Each of these makes a batch slower or smaller. They are on this page because they are the reason a batch is usable.
- Publish the number that could embarrass us
- Sync error, reject rates and thin coverage are on the site because a buyer finds all three in week one anyway.
- Pay for time, never per task
- A bounty per episode optimises for speed and produces footage that fails QA. We pay hourly, including setup and breaks.
- Discard rather than downgrade
- If a bystander cannot be de-identified, or a session drifts out of calibration, the footage is deleted. It is not sold at a discount.
- Say where the alternative wins
- Ego4D is broader. EgoScale is larger. EgoDex has excellent hand pose. We publish that alongside our own numbers.
- No capture without a spec
- Unspecified hours are the reason teams throw away most of the footage they buy. We will not sell them.
What we will not do
The list matters more than the mission statement.
- Real patient footage
- Clinical capture happens only in training environments with simulated patients.
- Covert capture
- Every person in a manipulation region has signed a release. No hidden rigs, ever.
- Children
- No participants under 18, and clips with children in frame are discarded rather than blurred.
- Scraped footage
- We do not resell web video. Every hour in the catalog was captured by an operator we paid.
- Surveillance buyers
- We decline programs whose stated purpose is identifying or tracking individuals.
Case study
A bimanual loading policy that kept failing on half-full racks.
Anonymised at the buyer's request: a US humanoid lab, Series B, training a bimanual dishwasher-loading policy. Their footage was clean and their policy still failed the moment a rack was partially occupied.
- 620hours
- Validated hours delivered in 11 weeks
- 11skills
- Named skills in the spec, all cells filled
- +21pts
- Success rate on their internal dishwasher eval
- 7.4%
- Batch reject rate — recaptured at our cost
The lab had 4,000 hours of public egocentric footage and a policy that scored well in simulation. In the real world it stalled on any rack that was already half loaded, because almost nothing in the public corpora shows a person recovering from a bad grasp inside a cluttered fixture.
We wrote a spec with a hard quota: 15% of every batch had to be a failure and recovery — dropped plate, wrong slot, tine collision — with the recovery captured through to completion. That quota is the whole case study.
Their words, lightly trimmed: “The failure cases were the only part we couldn’t buy anywhere else, and they moved the eval more than the other 500 hours combined.”
Engagement timeline
- WEEK 0Spec written together: 11 skills, 6 kitchens, explicit failure-case quota of 15%
- WEEK 1Rig calibration on their object set — 34 SKUs of real crockery shipped to us
- WEEK 2First 50 h batch delivered in RLDS. They found a wrist extrinsics error. We recaptured 12 h.
- WEEK 4Cadence at 70 h/week. Coverage report per batch with the reject log attached.
- WEEK 11620 h delivered. Failure cases turned out to be the highest-value slice.
Anonymised with permission. Reference call available under NDA once a spec is scoped.
Behind Firsthand
Built by CloudPano.
Firsthand is the embodied-AI arm of CloudPano, a spatial-data software company. Collecting real-world capture at scale, and turning it into structured data teams can actually build on, has been our core business long before robots needed it.
CloudPano started in immersive 360° imaging: software that lets anyone capture a physical space and publish it as an interactive virtual tour. That grew into a portfolio of capture tools — AI-generated floor plans, LiDAR scanning apps, and 3D reconstruction — used by creators, agents, and operators in more than two hundred countries.
The thread through all of it is data. From the beginning we have run two motions side by side: custom collection, where we design and operate a capture program to a client's exact spec, and off-the-shelf data, where we package what we have collected into products anyone can license. Doing both taught us how to scale capture operations globally without losing provenance or quality control.
Firsthand applies that same discipline to egocentric video for embodied AI. Same operating principles, same consent chain, same refusal to sell unspecified hours — pointed at the hardest capture problem in robotics.
- Parent
- CloudPano — spatial capture & data software
- Reach
- Products used by creators in 200+ countries
- Portfolio
- 360° virtual tours, AI floor plans, LiDAR capture apps
- Primary service
- Custom collection and off-the-shelf data, at scale
- Firsthand’s role
- The embodied-AI data arm of the portfolio
Team
Who you actually work with.
Firsthand is built by the CloudPano team — the same people who have run global spatial-capture operations and software for years, headquartered in Houston, Texas and working from around the world.
- Zach CalhoonCo-FounderLeads sales and marketing for CloudPano.
- Sevka PodlipentsevProduct ManagerDevelopment team lead and manager.
- Michael KlopotSenior DeveloperCloudPano Live lead developer.
- Eugene SklyarovSenior DeveloperTeams and Enterprise developer.
- Anna NeverovaSenior DeveloperGoogle Street View integrations.
- Michael KonyukohvQA EngineerIntegration and user testing.
- John OniaCustomer Success LeadSupport leader and front-office.
- Jade CaymoCustomer SuccessSupport and vendor relationships.
- Robert GearInbound SalesNorth America inbound sales.
- Ali NawazOutbound SalesBusiness development team.
- Fawad ButtOutbound SalesBusiness development team.
Company facts
The boring details.
- Entity
- Firsthand — a CloudPano company
- Founded
- 2024
- Headquarters
- Austin, Texas, USA
- Operating model
- In-house capture ops, contracted operators, no capture subcontractors
- Team
- Robotics data engineering, capture operations, annotation review
- Security
- SOC 2 Type II in audit; delivery over signed URLs, per-org isolation
- Contact
- support@iFirsthand.com
We are a small team by design. Capture operations, QA review and data engineering sit in the same standup, because a reject reason that never reaches the engineer who wrote the ingest check becomes a recurring defect.
If you are an embodied AI team, the fastest way to evaluate us is the free sample pack, then a Spec Pilot against one named skill. If you run capture operations of your own and want to contribute hours, the standards we hold contributors to are published in full.
Press, partnership and research enquiries all reach the same inbox: support@iFirsthand.com.
Start with the sample, not a call.
Download the pack, read the schema, then bring us a skill you cannot currently train.