Design Partner Program

We are not looking for logos. We are looking for engineers who will run ElectriPy in production, tell us what breaks, and help us build the infrastructure layer that AI systems actually need.

Real telemetry

Anonymized latency, error rates, circuit-breaker trip frequency, token costs. No PII, no business data — just infrastructure signals that tell us where the layers need hardening.

Honest feedback

What broke. What was confusing. What API you wish existed. We are not looking for compliments — we are looking for the friction that kills adoption.

A real workload

Toy projects don't reveal real problems. We need partners running ElectriPy against actual user traffic, even at low volume, where real failure modes emerge.

Partner tiers

Founding Partner

3 available

Free for 6 months

then negotiated enterprise pricing

What you get

  • Direct Slack channel with core team
  • Weekly 30-min architecture call
  • Your use case shapes the roadmap
  • Named in documentation and release notes
  • Private beta access — all features before public release
  • Production incident response SLA: 4 hours
  • Custom policy pack authoring support
  • Co-author a public case study (optional)

What we ask

  • Deploy to real production workload (any scale)
  • Monthly structured feedback session
  • Share anonymized latency + error-rate telemetry
  • Honest public or private case study after 90 days

Technical Partner

10 available

Free for 3 months

then standard pricing

What you get

  • Private Discord access
  • Monthly architecture office hours
  • Feature request priority queue
  • Pre-release access to new layers
  • Production incident response SLA: 24 hours

What we ask

  • Deploy to at least a staging environment
  • Quarterly feedback survey
  • Share high-level usage metrics

Who we are looking for

  • AI-native startups scaling from prototype to production
  • Platform / infra teams at mid-to-large companies evaluating AI infrastructure
  • ML engineers frustrated with bolt-on observability and governance
  • Teams that have been burned by LLM provider outages or unexpected cost spikes
  • Companies with compliance or data-residency requirements for AI workloads

What telemetry actually means

We will never touch your prompts, responses, or user data. The signals we care about:

Circuit breaker trip rateHow often is your upstream failing?
Policy action distributionHow many requests are being sanitized vs. blocked?
Span latency by layerWhere is the runtime adding overhead?
Token cost by labelWhich features are driving your AI spend?
Retry attempt countsHow often are transient failures recovered?
Error type distributionRate limits vs. timeouts vs. model errors

Apply to the program

Drop your email and a sentence about your workload. We will respond within 48 hours. No sales call required to start — we will send you the repo access immediately.

Or email directly: matt.vegas@inference-stack.com