Design partner opportunity
ERAIS Gateway
Help us test a local model endpoint your team can control. The work needs to show who can use it, what it can handle and how to restore or roll it back.
Your starting point
Would this help?
This is a fit when
- A technically capable partner with one local-model endpoint decision
- A controlled environment where operational and failure drills are permitted
When another approach is a better fit
- A public production endpoint or 24/7 SLA
- A requirement for broad API compatibility or unrestricted model support
What we’re planning
Proposed outcomes.
- 01Supported configuration and deployment/configuration record
- 02Access, request, data, audit, and failure boundary model
- 03Accepted endpoint contract and operator runbook
- 04Monitoring, capacity, drain, restart, restore, canary, and rollback evidence
- 05Limitations register and witnessed design-partner decision
Where this stands
Built so far. What comes next.
No delivery contract until an exact candidate passes security, authentication/denied paths, load/admission, 24-hour soak, graceful restart, restore, rollback, artifact trust, and independent-operator checks.
Existing basis
- A developed local inference endpoint and lifecycle control surface
- Bounded request, identity, limits, deadlines, observability, artifact, and failure controls
- Operational tooling for preflight, drain, restart, restore, and rollback testing
Evidence still needed
- Exact-candidate external security and denied-path closure
- Representative load, admission, and at least 24-hour soak evidence
- Partner-environment restart, restore, rollback, and independent-operator completion
Working together / 6–10 weeks after gate
How we’ll work together.
Discuss one workload and the controls it needs. A delivery agreement waits until the exact candidate passes the release checks listed below.
The project steps in detail
- 01Discovery only until the gate receipt is signed.
- 02Freeze one workload, contract, environment, and acceptance matrix.
- 03Configure and test the bounded endpoint and operational controls.
- 04Run witnessed capacity, failure, restore, and independent-operator checks.
- 05Promote, remediate, or stop; no automatic production handover.
What we need before work starts
Before starting
- One approved local-model workload and customer-controlled environment
- Named infrastructure, security, workload, and decision owners
- Supported hardware/runtime and lawful model/data use
- Permission to run failure, load, restore, and rollback drills
- Acceptance of a pre-production, one-configuration support boundary
Scope and handover
What we agree before starting.
We agree the work in a signed statement of work before starting. This page explains the offer; the signed agreement sets the commitment.
How you accept the handover
- One supported configuration passes the signed endpoint, identity, boundary, quota, deadline, and failure tests.
- Capacity, monitoring, graceful operation, restore, and rollback are witnessed in the partner environment.
- A named independent operator follows the runbook without founder-only knowledge.
- Limitations and support boundaries are accepted before any operational handover.
What this work covers
Discovery and non-binding design work only; no delivery contract until the pre-pilot gate passes.
One approved workload and supported configuration. No OpenAI-wide compatibility, general availability, production SLA, or customer-operability claim before the gate passes.
Implementation breadth is not production reliability, full API compatibility, external operability, or a support commitment.
Offer source and review record
Talk to us
What’s getting in the way?
Tell us what keeps going wrong and what you need to do once it’s sorted. We’ll come back with what we think could help.