Inside the talent bot: from CV to profile
A look at what happens after a candidate sends a PDF to the MellaWork talent bot.
Liya Mekonnen
MellaWork editorial
The talent bot exists for one reason: applying should not require a laptop and a quiet evening.
Send a PDF
A candidate sends a CV to the talent bot. The file is stored privately, hashed and queued. Nothing is parsed in the chat process; slow work belongs in the worker.
Extract structure
The worker reads the document with Docling, then asks a local model for a structured profile: roles, dates, education, skills and links. The candidate reviews what was extracted before it becomes their active profile.
Match to ESCO
Skills and occupations are mapped to ESCO concepts. When a candidate edits a skill, the change is saved immediately; the embedding refresh happens in the background.
One profile, many applications
The reviewed profile is reused for applications. A CV is evidence, not a separate identity. Employers see the snapshot the candidate submits, and the candidate can see exactly what was shared.
The result should feel like sending a message, not filling a form.
Hiring, simplified