AI API data residency in India

Data stays in India — DPDP-aligned, rupee-billed

unoblox's own infrastructure — the gateway, logs, and self-hosted models — runs on Indian data centers (Yotta, Haryana). Your API requests, logs, and inference on unoblox-hosted models stay on Indian servers by default. Billing is rupee-only, invoiced from an Indian entity, with no cross-border payment step involved.

How a request flows

Your app (India)
      |  HTTPS -> api.unoblox.ai
      v
Gateway layer (Yotta, Haryana)
  - auth, rate limiting, logging
      v
Inference
  - unoblox-hosted models (Qwen3, embeddings, reranker): served on Indian infrastructure
  - upstream provider models (GPT, Claude, and others): routed through the same gateway,
    key and billing stay on unoblox

What "data residency" covers here

Area	What happens
Your API key	Never leaves unoblox's systems; upstream providers never see it
Request logs	Stored on Indian infrastructure, encrypted at rest
Billing	Rupee-only, invoiced from an Indian entity, no cross-border payment step
unoblox-hosted models	Inference runs entirely on Indian infrastructure (Qwen3-1.7B, embeddings, reranker)
Upstream provider models	The request itself is routed to that provider's infrastructure to generate a response — this is inherent to using GPT, Claude, or any model unoblox doesn't host itself

Data lifecycle

1. Request arrives at the gateway and is logged (encrypted at rest, India-hosted).
2. Inference runs — on unoblox's own infrastructure for self-hosted models, or routed to the relevant upstream provider for models like GPT or Claude.
3. Response returns to your app; request/response logs are retained for a bounded period, then deleted.
4. You can export your own activity on demand for your compliance records.

Being straightforward about upstream models

If you call a model unoblox doesn't host itself — GPT, Claude, DeepSeek, and others — that request necessarily reaches that provider's infrastructure to run, the same as it would if you called them directly. What unoblox changes:

• Billing stays entirely rupee-denominated, invoiced by unoblox, regardless of which provider served the request.
• Your account credentials stay with unoblox — the upstream provider sees the request, not your unoblox login or user identity.
• Logs on unoblox's side stay in India and are available to you for audit.

If full data isolation to Indian-only infrastructure is a hard requirement, restrict your usage to unoblox-hosted models (Qwen3-1.7B, Qwen3-Embedding, Qwen3-Reranker) and check each other model's page before adopting it for a workload with that constraint.

Frequently asked questions

Does my data ever leave India? For unoblox-hosted models, no — inference runs entirely on Indian infrastructure. For upstream provider models (GPT, Claude, etc.), the request reaches that provider's infrastructure to generate a response, the same as calling them directly would.

Who are unoblox's sub-processors? Yotta (India) hosts unoblox's own infrastructure and is named in unoblox's data-processing documentation.

Can I request deletion of my data? Yes — contact support with a deletion request; it's honored within a bounded window except where compliance holds require retention.

Do you sell or share usage data? No. Aggregated, non-identifying analytics (which models are popular, overall volume trends) may be published; individual request content and PII are not shared.

Is there an audit trail I can export? Yes — your account's activity log is exportable on demand for your own compliance records.

How does this relate to DPDP? unoblox's own data handling — storage location, retention, and consent at signup — is designed around DPDP's requirements. For upstream provider models, review that provider's own terms as you would for any third-party API call.

Get started in rupees → https://unoblox.ai/sign-in