JARVIS Career Agent Privacy Policy

Effective date: September 27, 2026

Operator: David Russell Meredith — The Meredith Collective

Location: DTC, Colorado, United States

Privacy contact: administrative-agent@tmcsolutions-org.net

Purpose and current availability

JARVIS Career Agent is a private, single-owner career-management application intended to help its owner find steady employment and income. It organizes career evidence, job postings, resumes, applications, and follow-up activity. It is not offered for public registration or use by other people.

The application is still being provisioned. LinkedIn sign-in, Google Gmail, Calendar, and Drive integration, and AI-assisted resume tailoring are intended for the initial release. This notice describes implemented data paths and clearly identifies planned functionality; publishing it does not mean the application or every integration is operational. Databricks application permissions, not LinkedIn sign-in, are intended to restrict access to the owner.

Career information

The application processes owner-supplied employment history, education, certifications, skills, and other career evidence. Job-tracking records can contain company and role names, job URLs, locations, compensation ranges, application statuses, dates, and private notes.

The configured database stores postings, application records, resume-version metadata, match scores, gap analyses, and application activity. Generated resume text is returned to the browser. Saving a separate resume file depends on the storage integration. These records support the owner’s career preparation and application tracking, not advertising or data brokerage.

LinkedIn sign-in is included in the launch plan but remains subject to setup and live verification. When the owner chooses to connect, LinkedIn handles authentication and consent. The application requests openid, profile, and email. It does not collect the owner’s LinkedIn password.

The sign-in callback exchanges an authorization code for an access token and requests LinkedIn’s user-information response. That response may include a member identifier, name, profile-picture URL, email address, and related identity claims, depending on what LinkedIn returns.

The implemented callback retains only the member identifier and name in a signed browser session cookie to identify the connected account. It processes the token and identity response in server memory and does not intentionally persist the LinkedIn access token, email, or picture in the application database or cookie. This is not a guarantee about provider diagnostic logs.

Identity is retrieved during sign-in, not by scheduled background refresh. This integration does not access LinkedIn messages, connections, or job-search history. The LinkedIn session identity is not an input to resume tailoring. Separately supplied career evidence is distinct from identity information retrieved through this sign-in integration.

Google integrations

Gmail, Calendar, and Drive are core requirements, not merely optional future enhancements. Durable Google authorization, token refresh, and the production Drive writer are not yet complete. They must be implemented and verified before the owner relies on them for ongoing use.

When configured and invoked, the existing Gmail/Calendar synchronization code requests recent Gmail message metadata and upcoming primary-calendar events. It can process message identifiers, sender headers, subjects, timestamps, snippets returned by Gmail, event titles, and attendee email addresses to match activity to applications. Matching occurs after retrieval; the initial retrieval is not restricted to known recruiters. Matched subjects or event titles, source identifiers, and timestamps can be stored as application activity in Lakebase.

The reviewed Gmail path does not request full message bodies or attachments, although the intended gmail.readonly authorization permits broader reading than these requests. Calendar access is intended to be read-only. These paths do not send emails or change calendar events, and their implemented matching logic does not call an AI model.

The planned Drive integration will save generated resume artifacts in the owner’s selected folder and retain file identifiers with resume-version records. The current route exposes a writer extension point, not a verified production Drive implementation. Only permissions necessary for the implemented feature will be requested; exact access and any additional data flows will be disclosed before authorization.

Any enabled use of Google API data, including transfers, must follow the Google API Services User Data Policy, including its Limited Use requirements. Google data will be used for the disclosed owner-facing features, not advertising, data brokerage, or training generalized AI models. Additional access or use requires policy review and any required consent before activation. This is an operating commitment, not a claim of Google approval or a completed security assessment.

AI-assisted resume tailoring

When the owner requests tailoring and the feature is configured, selected posting fields and career evidence are sent to Anthropic’s API. The evidence includes relevant experience, education, certifications, and skills. Anthropic returns resume text, a match score, keyword analysis, and suggested certifications. These inputs can contain personal information such as employment history.

The implemented tailoring path does not read applications.notes, LinkedIn session identity, or Gmail/Calendar activity as prompt inputs. This field separation does not remove sensitive information independently placed in career evidence or other AI inputs. The owner should submit only information they are willing and authorized to provide for AI processing.

Outputs require human review. The reviewed tailoring route does not automatically submit applications or send resumes to employers. Anthropic’s applicable service terms and account settings govern its handling of API inputs and outputs; this policy does not promise zero provider retention or a particular processing country. No assertion is made that the operator has a special zero-retention agreement.

Providers and access

Databricks and its infrastructure providers: hosting, workspace authentication, Lakebase database storage, and operational services. Authorized platform administrators may have access under the configured permissions.
LinkedIn: authentication, consent, and the identity response described above.
Google: authorized Gmail, Calendar, and Drive operations when enabled.
Anthropic: the AI processing described above when requested and configured.
Cloudflare: hosting this public documentation site on Cloudflare Workers. Cloudflare processes visitors’ IP addresses and request metadata to deliver and protect the site, as described in its privacy policy. This site contains public policy documentation, not private career records or application credentials.
GitHub: hosting the documentation source repository. Until the previous GitHub Pages copy of this site is retired, GitHub Pages also serves it and logs visitors’ IP addresses for security, as described in its Pages documentation.

The operator does not sell personal information, use it for advertising, or provide it to data brokers. No additional provider is currently designated. Adding providers requires review of their data access, an updated notice, owner approval, and any required consent; this policy is not blanket authorization for unspecified future sharing.

Provider processing locations, service logs, and backups are governed by applicable provider terms and settings. The owner-managed career-record schedule below is not a guarantee that provider copies have the same lifetime or remain in one country.

Cookies and logs

The application uses a signed session cookie for LinkedIn OAuth state and the connected identifier and name. A signature protects against undetected modification; it does not encrypt the cookie’s contents. Clearing application cookies removes that browser’s stored session but does not revoke provider authorization or delete database records. Browser session restoration can preserve session cookies.

Databricks and authentication providers can use their own cookies and process connection metadata, requests, errors, and identifiers. Cookie configuration, log content, and provider retention must be checked before deployment; no fixed deletion interval for logs or backups is promised by this notice. Additional Unity Catalog telemetry export has not been verified as enabled. This documentation site does not add advertising trackers or analytics scripts.

Retention, archives, and deletion

The owner has adopted these requirements for owner-managed career records:

Application records: remain active while an application is open and for one year after application closure, then are archived and compressed.
Archives: remain until the owner explicitly approves deletion, without a fixed automatic expiry. Archives remain personal information; compression is not deletion or anonymization.
Master career evidence: employment history, education, and certifications remain active indefinitely, separate from the one-year application-record schedule.

The archival schedule is not yet automated in the reviewed application. Records currently remain until the operator acts on them. Closure-date tracking, archival, and deletion procedures still require implementation and verification. There is no claim that a scheduled process already archives or erases data.

David Russell Meredith (Russ Meredith / Russypher) is responsible for administration and privacy requests. J.A.R.V.I.S. may assist with authorized administration, but every deletion by J.A.R.V.I.S. requires the owner’s explicit permission. Permission to archive or compress is not permission to destroy source records. J.A.R.V.I.S. is not a separate legal operator, and this policy itself grants no technical privileges.

This career-record schedule does not authorize keeping provider-sourced data or credentials contrary to applicable law or provider requirements. Applicable deletion obligations take precedence, including LinkedIn’s requirements for deletion of API data on request, account closure, or other specified events. The owner remains responsible for approving and completing required actions promptly.

Choices and requests

Contact administrative-agent@tmcsolutions-org.net to ask about access, correction, export, or deletion. Identify the relevant records without sending passwords, API keys, or tokens. The owner handles requests, may seek proportionate confirmation of identity, and authorizes any necessary administrative actions.

LinkedIn and Google authorization can be withdrawn using the respective account’s connected-application controls. Revocation stops future authorized access but does not automatically delete already-stored records or application cookies. Request removal separately. The reviewed app does not yet provide a dedicated account-deletion or LinkedIn disconnect/logout endpoint.

The owner can choose not to connect a provider or invoke AI features. Provider-controlled records and backups may require separate provider procedures; instantaneous deletion of every copy is not promised. Mandatory legal or provider obligations continue to apply.

Security and policy maintenance

The intended deployment uses HTTPS, owner-restricted access, protected credentials, and a dedicated database role. Some controls still require deployment verification; no system or policy can guarantee absolute security. The application is not intended for children or a public user population.

This policy is maintained in the public documentation repository. The operator will review and update it before materially changing integrations, data uses, or the audience, and obtain additional consent where required. Applications and integration registrations using this policy should link to this current version. Version history records published changes. Before allowing other users, the operator will separately review access controls, privacy obligations, and change notifications.

Privacy questions: administrative-agent@tmcsolutions-org.net.