HR Helper: AI scoring of candidates with an explanation of each score

In hiring, it is difficult to quickly and fairly compare candidates: resumes come in different formats, and part of the decisions is based on feelings. I have created a system that reads resumes, breaks down job vacancies into criteria, and filters out those who clearly do not fit. Next to each score is a quote from the resume and a reliability tag, so the assessment can be verified rather than just trusting the model.

What’s inside:
- Acceptance of resumes in PDF, DOCX, HTML, Markdown, text, and complex scans, even when regular text extraction does not work.
- Job profile with criteria in three groups: skills, experience, and stop-factors. Each criterion has its importance, with versions and history for calibration.
- Detailed assessment: each criterion receives a score and a quote, and the system separately generates an overall conclusion with strengths and risks.
- Multiple model mode: they assess the candidate independently, and the system marks any disputes.
- Mini-CRM funnel with 8 statuses: interview question form, draft rejection letter, interview transcript, dashboard, and calendar.

17 tables, 60 routes, 18 prompts. What should be predictable remains in the code, so the result can be verified.

#Python #FastAPI #AI #HRTech #LLM #Recruiting #Automation #SQLModel #SQLite #HTMX #Tailwind #PyMuPDF #MarkItDown #OpenRouter #Whisper