Subprocessors

Boardwalk Compute Company uses the providers below to operate Boardwalk. This is the canonical list incorporated into Annex III of our Data Processing Addendum. A provider is listed when it may process Customer Personal Data on our behalf, even if it receives only a narrow category of information.

Amazon Web Services, Inc.

• Purpose: primary cloud infrastructure, including API compute, model ingestion, deploy orchestration, the isolated function that converts a repository’s own tokenizer code, secret storage, email delivery, queues, scheduling, monitoring, and the independent status service.
• Data: account and organization records, model artifacts and metadata, transient synchronous inference content, explicit batch input and result files, usage and billing records, communications, secrets, and operational logs as needed for the relevant AWS service.
• Location: United States, primarily us-east-1, subject to AWS support and service operations.

Shadeform, Inc.

• Purpose: the marketplace through which we rent dedicated GPU machines that run Boardwalk's serving software for deployed models. Each machine is operated by a GPU cloud provider that Shadeform engages, chosen per rental from those with capacity in the United States.
• Data: inference prompts and outputs in the machine's memory during request processing; copies of the model artifacts a deployment serves, including private ones; and technical logs and metrics produced on the machine.
• Location: United States only. We rent a machine only when its listed region is in the United States.

Runpod Inc.

• Purpose: serverless GPU compute for catalog models served from shared pools, and the network volume used for their model and engine artifacts.
• Data: inference prompts and outputs during request processing; request and model identifiers; public, private, or gated model artifacts needed to serve a deployment; and technical logs and metrics produced by workers.
• Location: United States; Boardwalk's current shared cache volume is in RunPod's US-IL-1 data center. Worker placement is constrained by the serving configuration and RunPod capacity.

Cloudflare, Inc.

• Purpose: object storage (R2) for the copies of model artifacts that GPU machines load from, and for the serving software those machines fetch.
• Data: model artifacts, including private weights, configuration, and tokenizer files. No inference content is stored there.
• Location: United States only. Our buckets are in R2's United States jurisdiction, which keeps their objects in the United States.

Tailscale Inc.

• Purpose: the private network between Boardwalk's backend and the GPU machines it rents.
• Data: machine names, network addresses, and connection metadata used to coordinate the network. Inference traffic between our backend and a machine is end-to-end encrypted; where it passes through a Tailscale relay, it does so only in encrypted form.
• Location: United States, with relay servers in other locations.

PlanetScale, Inc.

• Purpose: managed PostgreSQL database for the Boardwalk control plane.
• Data: account and organization records, API-key metadata, model and deploy metadata, token and cache counts, request receipts, ledger and payment records, invitations, and audit events. Prompt and completion text is not stored.
• Location: United States.

Clerk, Inc.

• Purpose: authentication, account identity, session management, profile management, and multi-factor authentication.
• Data: email address, display name, profile image, authentication factors, device and session information, and authentication metadata.
• Location: United States, with service delivery and support operations as described by Clerk.

Stripe, Inc.

• Purpose: checkout, payment processing, saved payment methods, auto top-up, refunds, disputes, fraud prevention, and related billing operations.
• Data: billing contact and organization information, purchase amounts, payment method and transaction information, and fraud-prevention signals. Full card details are submitted directly to Stripe.
• Location: United States and other locations in Stripe's global payment network.

Vercel Inc.

• Purpose: hosting and delivery of boardwalk.cloud, app.boardwalk.cloud, admin.robotnet.works, and status.boardwalk.cloud.
• Data: frontend requests, IP address, user agent, session cookies and tokens in transit, account and organization data rendered or submitted through the console, and contact-form submissions.
• Location: United States, with edge delivery points globally.

PostHog, Inc.

• Purpose: website and product analytics.
• Data: pseudonymous browser identifiers, IP address, approximate location, device and browser metadata, page views and interactions, and—when signed in—Boardwalk user and organization identifiers and names. Selected server-side product outcomes include internal identifiers, timestamps, status, and limited numerical or categorical metadata independently of browser cookie preferences. We do not send email addresses, API keys, payment-card details, prompts, or completions. Session replay is disabled.
• Location: United States.

Customer-directed providers

Hugging Face is a service Customer chooses to connect. We exchange OAuth and repository information with Hugging Face under Customer's account and its terms. We do not treat Hugging Face as our Subprocessor when it acts under that customer-directed relationship. Third-party repositories and model licensors likewise remain subject to their own terms.

Changes and objections

We will post an update and give DPA customers at least 30 days' notice before a new or replacement Subprocessor begins processing Customer Personal Data. Subscribe by emailing legal@boardwalk.cloud with the subject "Subprocessor notifications". Objections are handled under Section 6 of the DPA.

Change log

• 2026-09-05: initial Boardwalk inference-service list.

Contact

Questions: legal@boardwalk.cloud.