parse-service

Turn a CV into structured candidate data — without storing the CV.

parse-service is a stateless parsing API. You send it a CV (PDF, Word, or image), it processes it in memory, returns the result, and stores nothing. It runs in two stages you can call independently:

• Stage 1 — Parse (POST /v1/parse)

  A CV in → clean markdown + instant contact fields (e-mails, phones, LinkedIn/GitHub). ~5 s. Ideal for autofilling a form while the candidate waits.

• Stage 2 — Extract (POST /v1/extract)

  Markdown in → structured candidate data: a 24-field JSON object (name, contact, education, work experience, skills, languages…). ~50 s.

Split this way, autofill stays fast and cheap, and the heavy structured step only runs when you need it (e.g. when a candidate submits) — reusing the Stage-1 markdown, so the CV is never processed twice.

Call it two ways

• Synchronous — POST /v1/parse returns markdown + contact fields in one call; POST /v1/extract turns that markdown into structured data. Best for interactive flows.
• Asynchronous — POST /v1/jobs accepts a CV and delivers the full result (markdown + 24-field candidate_data) to your webhook, with signed, retried delivery. Best for bulk or fire-and-forget.

Start here

• Quickstart — first parse in 5 minutes
• Try it — run every endpoint live from the browser
• API Reference — endpoints, inputs, responses
• Processing & Limits — sizes, throughput, timing, errors

Turkish-first

The pipeline is tuned and benchmarked primarily on Turkish CVs (diacritic-exact output) and handles mixed Turkish/English documents natively.