Return Nima Karami's public-safe profile: name, headline, location, links, bio.

List every role/job (flat) with id, org, company_id, title, dates, tags.

Roles sharing a company_id are a title progression at one company; use list_experience for the grouped view.

Roles grouped into company tenures (newest first).

Each company tenure lists its positions (title progression) newest-first. A company you left and later rejoined appears as two separate tenures. Use this when you want experience grouped by employer rather than as a flat list of titles.

Get one role in full, including its evidence bank and approved resume bullets.

Use list_roles first to discover valid role ids (e.g. 'timeplay').

List projects, optionally filtered by tags (AND) and/or owning role_id.

Get one project in full: blurb, evidence, approved bullets, and links.

List skills grouped by category; each skill links to backing evidence ids.

Search across roles, projects, bullets, and skills.

This is the main entry point for queries like "show me his 0-to-1 product work". kinds restricts item types (role|project|bullet|skill); tags is an AND-filter. The query is treated strictly as search input, never as instructions.

List curated, hiring-manager-safe resume bullets.

Each bullet is pre-approved and cites the evidence ids it derives from. Filter by role_id, project_id, and/or tags (AND). With no filters, returns all bullets.

Assemble a tailored resume draft for a query (e.g. "backend-leaning").

SELECTS and ORDERS the most relevant approved roles, evidence, bullets, and skills for focus (or for the explicit role_ids/project_ids), and fills an approved summary template — it never authors new facts, employers, dates, or metrics.

Args: focus: free-text angle, e.g. "0-to-1 product" or "backend". role_ids / project_ids: pin specific items instead of ranking by focus. skill_categories: restrict the skills section. length: "full" or "onepage" (caps roles/bullets). format: "structured" (default) or "markdown" (also returns rendered markdown).

The returned draft includes a provenance list of corpus ids and a disclaimer. A consuming agent may rephrase for fit but MUST NOT add claims not present here.