An engineer uses a laptop inside a mechanical equipment frame with visible wiring and components.

Research Pilots

A research pilot is a bounded study of whether a proposed method can support a defined industrial task. Its purpose is to resolve a specific uncertainty using evidence from an agreed setting. A pilot should have a clear question, a meaningful comparison, and an end decision that remains useful even when the candidate does not improve on the baseline.

Use this page to discuss a pilot with Aperture Physical AI Research Labs. Describe the physical system, the method or question under consideration, the observations available, and the work the result could influence. Scope, fees, schedule, and engagement terms are agreed for each project after its requirements are understood.

Bound The Task and The Claim

Identify the equipment, site, users, operating conditions, and period included in the proposed study. State what the method estimates or recommends and what it may influence. A recommendation for human review differs from an output connected to a physical command.

For example, a study might examine whether condition observations help a maintenance team prioritize inspections on a defined equipment group. That question does not establish autonomous diagnosis, authority to change equipment settings, or performance across other sites. The scope should preserve those distinctions.

Agree The Baseline and Reference Evidence

Choose a relevant existing procedure or justified simpler method for comparison. Explain what information each alternative receives and when it becomes available. Compare matched conditions rather than giving the candidate information that the baseline could not have known at the decision time.

Define how outcomes will be established independently of the candidate. Keep uncertain findings visible, including incomplete inspections and unresolved equipment associations. Development examples and final evaluation cases need an appropriate separation so repeated tuning does not become its own evidence of success.

Match Access To The Stage Of The Study

Recorded and digital evaluation

Recorded observations can test a method against defined cases. Simulation can examine selected assumptions and behaviors. State which evidence is recorded, simulated, or physically measured; the settings establish different claims and have different limitations.

Live shadow and supervised work

Live shadow operation can examine current inputs, timing, and outputs without allowing the candidate to change the task. An authorized supervised pilot can study effects on the actual workflow within an agreed boundary. Specify the stage proposed and the permissions needed for that stage.

Define the response to an unsuitable condition

Agree who can pause or end the study and how work returns to the established procedure. Missing observations, unavailable supervision, a changed equipment configuration, or an unresolved output may require a pause. These conditions should be part of the design rather than improvised during a trial.

Assign Supervision and Responsibilities

Name the research lead, operational contact, equipment authority, technical reviewer, and person responsible for the next decision. Define who prepares inputs, resolves ambiguous cases, records outcomes, and approves scope changes. Include the practical effort of those activities when planning resources.

Research participation does not replace the site's engineered control and protective arrangements. Identify permitted interfaces and preserve the distinction between analytical review and authorized physical intervention. Access to software or data alone does not establish permission to change equipment behavior.

Evaluate The Complete Workflow

Choose measures that follow the task, including relevant errors, coverage, unresolved cases, timing, and review effort. For an inspection priority study, record whether recommendations were received, reviewed, and followed by a completed inspection. A generated alert is not evidence of a completed maintenance outcome.

Determine whether the observation period and number of independent events can support the intended conclusion. Report sparse outcomes and differences between operating groups. Aggregate results can conceal an important limitation, and a short study may leave rare conditions unresolved.

Agree The Record and Next Step


Before work begins, agree access, resources, permitted data use, deliverables, confidentiality, review, and relevant publication arrangements. Record how changes to the candidate or study will be handled. The project agreement defines the actual commitments; this page does not prescribe a standard price or duration.

At the end, document the scope, baseline, candidate version, reference evidence, findings, limitations, and continuing effort required. The next step may be another bounded experiment, a revision, a narrower use, or an end to the work. Broader deployment requires its own evidence and authorization.

Scroll to Top