Engagement Variants — apply the methodology to your documents

The two engagement variants

Both variants ship the same methodology, the same recipe, and the same Apache 2.0 license posture. The variant determines what is in the box.

• Inference-online engagement
  Your documents are processed via Anthropic Claude, Google Gemini, or Mistral APIs on Synavistra accounts during the engagement. You take home: the recipe (Apache 2.0), a complete log of every interaction with the AI, a signed provenance record, and instructions to set up your own API account if you want to keep running the analysis afterwards.
• Inference-offline engagement
  Everything in the inference-online variant, plus a customer-fine-tuned Gemma 4 model, the training-stack source, and the recipe to reproduce the full training infrastructure on your own cloud or hardware. You can run the methodology offline indefinitely, on your own hardware, air-gap-capable for sensitive data.
• Or: see the work first
  If you want to see the methodology in action before scoping an engagement, browse the Showcases for worked examples of Synavistra deliverable patterns. Each is built on a public source so the methodology is inspectable.

The unifying promise

Every engagement ships under the same commitment, regardless of variant:

• No kill switch.
• No license-expiry that disables your code.
• No annual relicensing fee.
• If you want to apply the recipe to other internal use cases, you may.
• If Synavistra disappears tomorrow, your stack keeps working.

What every variant ships

Three parts in every deliverable bundle. The customer takes home all three end-to-end.

• Recipe
  Apache 2.0 source code, infrastructure recipes, configuration, and the instructions that produced the deliverable. Permanent and irrevocable.
• Provenance manifest
  A signed record listing every artifact, its content fingerprint, its license, and (for inference-online) a complete log of every API call made on your behalf. Hand it to your auditor or regulator any time you need to demonstrate provenance.
• Customer artifacts
  The customer's processed documents, results, and (for inference-offline) fine-tuned models. Owned end-to-end by the customer under the customer's license; Synavistra has no claim and retains no encryption keys.

How an engagement runs

Both variants follow the same lifecycle. Synavistra runs the engagement on Synavistra's own infrastructure and accounts; the customer's only obligation at start is the signed Statement of Work.

• Phase 0 — Pre-engagement
  Scoping conversation, SOW drafting, deposit invoice.
• Phase 1 — Engagement work
  Synavistra runs the entire pipeline on Synavistra's accounts. The customer provides source documents (and, for inference-offline, fine-tuning data). Engagement-window API costs accrue to Synavistra and are absorbed in the engagement fee.
• Phase 2 — Delivery
  Single signed handover: a signed manifest plus an encrypted archive sent to a public key you provide. For inference-online, Synavistra closes its engagement-scoped API access at the moment of delivery — no further requests are possible from our side.
• Phase 3 — Walk-away
  Synavistra has no continuing role. The customer has every artifact under the three license rules. One delivery, one signed manifest, one walk-away — Synavistra ships, customer takes home, both walk away.

You own everything. No strings.

One delivery, one signed manifest. Apache 2.0 on Synavistra's work, full IP on yours, complete third-party license inventory. Take it home and never look back.

• Synavistra-authored material is Apache 2.0 — permanent, irrevocable, redistributable. No kill switch. No annual relicensing fee. The customer can apply it to other internal use cases.
• Customer data, results, and trained models are the customer's property under the customer's license. Synavistra has no claim, no encryption keys, no embedded restrictions.
• Third-party components retain their own licenses. Each one is listed by name + license + version + license-text URL in the per-engagement deliverable manifest.

Email partnerships@synavistra.ai with a brief description of the documents you want analysed and the question you want answered from them. We will reply within two business days with a scoping-call invitation if there is a fit.