Predicate and claims narrative
Substantial-equivalence strategy, intended use, comparison tables, technological differences and evidence needed to address them.
AI/ML submission
AI/ML 510(k) submission prepared in three months.
For teams with a locked model, available test data and a settled regulatory strategy, Predetics runs software, risk, cybersecurity, performance and submission work as one controlled sprint.
Best suited to
Locked AI/ML models
Teams with available validation data
Products with an agreed 510(k) strategy
Radiology and diagnostic software
Sponsors with committed engineers
Companies treating submission as a top priority
Scope and deliverables
What the submission includes.
The submission must let reviewers understand what the model does, how it was developed, where it can fail and why the validation supports the proposed use.
Software lifecycle file
Requirements, architecture, detailed design, traceability, configuration management, anomaly handling and verification records.
AI/ML evidence
Dataset provenance, independence, representativeness, ground truth, subgroup analysis, acceptance criteria and performance reports.
Clinical study package
Standalone, MRMC or other clinical-performance protocol and reporting support where required by the claims and strategy.
Risk and cybersecurity
Risk management, threat model, SBOM, security architecture, testing, vulnerability assessment and lifecycle plans.
Complete 510(k)
Administrative content, device description, software and performance sections, labeling, declarations, eSTAR assembly and filing.
Delivery plan
Three-month submission schedule.
- 01
Sprint 1 · Control
Baseline the model and software, close strategy decisions, define acceptance criteria and finalise verification and validation protocols.
- 02
Sprint 2 · Evidence
Execute or supervise testing while completing design, risk, usability, cybersecurity and traceability documentation.
- 03
Sprint 3 · Submit
Complete performance reports, predicate narrative, labeling and eSTAR, then run independent quality review and file.
- 04
FDA review · Respond
After the three-month filing, support interactive review and any Additional Information response under a separately governed FDA clock.
Guarantee boundaries
What the commitment covers.
- Three months means submission to FDA, not FDA clearance; FDA controls review and may issue questions or an Additional Information request.
- The algorithm must remain locked except for controlled corrections agreed through change control.
- Required datasets, readers, software access and client engineers must be available from programme start.
- Performance failures, scope changes and client-caused delays are handled under the written guarantee exclusions.
What we need from your team
What keeps the schedule on track.
- →Assign one or two engineers who can answer architecture, model and test-execution questions quickly.
- →Provide controlled source code or technical access, a reproducible build and a deployable model package.
- →Make independent validation data and required clinical readers available on schedule.
- →Freeze intended use, major claims and model performance targets for the submission sprint.
Regulations, standards and guidance considered
FDA 510(k)IEC 62304ISO 14971IEC 62366-1FDA AI-Enabled Device GuidanceFDA Cybersecurity GuidanceNIST SSDFIMDRF SaMD
Frequently asked questions
Before we begin.
Does three months mean FDA clearance in three months?+
No. The commitment is to submit the agreed package within three months. FDA’s review begins after submission and follows the Agency’s own timeline.
Can we qualify if the model is still changing?+
The accelerated programme requires a locked model and stable intended use. If development is ongoing, the concept-to-cleared-device programme is usually a better fit.
Do you run the clinical study?+
We can design protocols, acceptance criteria, analysis and reports and coordinate suitable partners. Reader recruitment, site operations or data acquisition may be separate.
Can you support a De Novo AI device?+
Yes, but a novel De Novo programme may require a different schedule because the risk-control and clinical evidence burden is often greater.