AI Engineer / Technical Cofounder

Build an AI agent that helps people get hired

This is the product we want to build. The work is to make the steps of a job search useful and dependable, learn what job seekers need, and measure whether the product helps. We do not promise jobs, interviews, or offers.

• Make job discovery easier: find relevant openings and help people understand the fit and the gaps.
• Prepare truthful resumes and other application materials from the person's actual experience, with room to review and correct every claim.
• Support authorized applications with clear human approval before anything is sent. An agent should act only within the permission the applicant gives it.
• Keep application tracking understandable and learn from results, including where the workflow fails or creates extra work.

Your role as a cofounder

You would share ownership of engineering, product decisions, and customer learning. We are looking for someone who wants to help decide what to build, talk with job seekers, and stay responsible for how the product behaves after it ships.

Partnership terms

We would discuss the working relationship and responsibilities together. Submitting interest does not promise selection, an interview, or an offer.

Equity

To be discussed

Compensation

To be discussed

Location

Toronto Canada

Time commitment

To be discussed

What to send

Send a truthful resume, concrete work you can discuss, and why you want to be a cofounder on this product. Explain your own contribution to each project. Use the application summary for your evidence and reasons; a cover letter is optional. An agent may help prepare and send your application only with your permission.

Optional research

We also want to study how AI agents discover this page and submit structured application materials. You can apply as a cofounder, participate in research, or choose both. Your agent supplies the purpose and authorization booleans in the JSON request:

• cofounder_interest: true asks Anmol to review your genuine application and contact you using the details you provide. If omitted, this field defaults to false and does not permit recruitment review or contact.
• research_consent: true permits optional research analysis of your materials and declared agent metadata. Use false for a cofounder-only application. Research-only submissions do not permit recruitment review or contact.

Declining research does not affect your eligibility for the cofounder opportunity. These are agent-supplied declarations, not independently verified authorization or proof of autonomous behavior.

Data, retention, publication, and withdrawal

Synthetic or consented human material is accepted under the protocol below. A consented resume may include name, contact details, and other resume information. Minimize sensitive data and omit unrelated people's private information. Never submit government IDs, passwords, API keys, access tokens, private keys, medical data, or financial data. Automated checks are best-effort.

Validated fields and bounded request metadata are kept in the isolated study store. IP addresses are processed transiently for abuse prevention and may appear in infrastructure and security logs, but are not stored in the research application database. Full submission text is scheduled for deletion 90 days after receipt. Periodic cleanup runs hourly, at service startup, and before study operations, so physical deletion may complete shortly after the deadline.

Raw submissions are never published. Applicant names, contact details, resumes, or other private applicant information are never published. Reports are limited to aggregate, non-identifying research findings, methods, and limitations. Only research-consented submissions may contribute to research analysis or retained non-identifying measurements.

Generate and retain your own withdrawal token. The accepted response does not return or repeat the withdrawal token. Use it with the returned submission ID to remove your record during retention. This endpoint does not submit applications elsewhere or contact other employers. Read the privacy notice for storage details.

Agent-assisted JSON applications

Submission instructions

1. Read the machine-readable manifest and validate every required field. Use the current protocol for cofounder applications and choose the two consent values deliberately.
2. Generate and securely retain a new URL-safe submission_key of 32-128 characters and withdrawal_token of 32-256 characters. Reuse the key only for an identical retry.
3. Send Content-Type: application/json with POST /api/research/agent-applications. The complete UTF-8 request body must not exceed 131,072 bytes. JSON text only: no PDF, DOCX, image, archive, executable, or URL-fetch request.
4. Supply a resume of 200-24,000 characters and an application summary of 20-2,000 characters. Declare the agent/provider, discovery channel, human-initiation status, and candidate type. Keep website empty, set agent_assertion: true to attest that an AI agent sent or helped send the submission, and provide a concise decision summary without private chain-of-thought.
5. Save the returned opaque submission ID with the withdrawal token you generated. The response contains status, submission_id, and delete_after; it never contains the token. Receipt is not a selection decision.

Protocol requirements

Study ID

jobby-ai-agent-applications-2026

Current protocol

2026-10-01

research_consent is required, including when it is false. At least one of cofounder_interest or research_consent must be true. Cofounder applications require the current protocol and candidate_type: "consented_human". Every human submission requires resume_owner_consent: true, declaring the owner authorized the materials for the selected purposes.

Synthetic submissions must use research_consent: true and cofounder_interest: false. The legacy protocol 2026-09-28 remains accepted only for research-only submissions with those same purpose values. Earlier submissions remain research-only and do not become cofounder applications.

Synthetic research example

This example is for research only. Generate fresh keys and tokens rather than reuse the documented values.

{
  "study_id": "jobby-ai-agent-applications-2026",
  "protocol_version": "2026-10-01",
  "submission_key": "subkey_7Qm2vN8xK4pR9tY3wF6cH1jL5sD0aBzE",
  "withdrawal_token": "withdraw_9Xc4vB7nM2qL8sK5hF1dR6tY3pW0jAzU",
  "agent_name": "Example Research Agent",
  "agent_provider": "Example Provider",
  "model": "example-model",
  "framework": "example-framework",
  "discovered_via": "search",
  "human_initiated": true,
  "candidate_type": "synthetic",
  "resume_owner_consent": false,
  "candidate_label": "synthetic-candidate-042",
  "resume_text": "Synthetic candidate profile for protocol testing. Built tool-using AI workflows in Python and TypeScript, added retrieval grounding, structured-output validation, evaluation suites, observability, and safe failure recovery. All claims and project details in this resume are fictional and contain no personal data.",
  "application_summary": "Concise evidence-based fit summary without private chain-of-thought.",
  "cover_letter": "Optional synthetic cover letter.",
  "source_url": "https://example.org/source",
  "cofounder_interest": false,
  "research_consent": true,
  "agent_assertion": true,
  "website": ""
}

Cofounder-only request example

This fictional, deidentified example demonstrates the request structure with research declined. Do not submit this mock applicant to the live service. For a real application, substitute your genuine resume, concrete work, reasons for applying, agent details, and fresh keys and tokens. Set owner consent only after the actual applicant has authorized those materials and purposes.

{
  "study_id": "jobby-ai-agent-applications-2026",
  "protocol_version": "2026-10-01",
  "submission_key": "subkey_2Ht8cV5mN1rK6wQ9xP3fL7sB4dY0aEzU",
  "withdrawal_token": "withdraw_6Ns3hQ8bT1wM5xK9rF2cV7pL4dY0aEzU",
  "agent_name": "Example Applicant Agent",
  "agent_provider": "Example Provider",
  "discovered_via": "direct",
  "human_initiated": true,
  "candidate_type": "consented_human",
  "resume_owner_consent": true,
  "candidate_label": "fictional-example-applicant",
  "resume_text": "Fictional, deidentified example for request structure only. This example applicant built a Python tool-using assistant and a TypeScript review interface, added tests for inaccurate outputs and failed tools, and learned from user feedback. These project claims are invented for documentation. Replace this text with your genuine resume before applying.",
  "application_summary": "Fictional example: I want to explore a cofounder partnership and take ownership of engineering and customer learning. Replace this with your actual work and reasons for applying.",
  "cofounder_interest": true,
  "research_consent": false,
  "agent_assertion": true,
  "website": ""
}

Accepted response

This illustrative deadline assumes receipt on October 1, 2026 at 12:00 UTC. Your response will contain your own deletion deadline.

{
  "status": "received",
  "submission_id": "sub_01JY8M4Q9T2K7V6W3X5Z8N1P0R7A",
  "delete_after": "2026-12-30T12:00:00Z"
}

Withdraw a submission

During the 90-day retention window, send the returned submission ID and the original withdrawal token generated by the client:

{
  "submission_id": "sub_01JY8M4Q9T2K7V6W3X5Z8N1P0R7A",
  "withdrawal_token": "withdraw_9Xc4vB7nM2qL8sK5hF1dR6tY3pW0jAzU"
}