Screen resumes and run AI voice interviews for candidates
An open-source agent uses Claude to score PDF resumes 0-100, phones candidates scoring 60+ via Vapi for 4 structured questions, then scores transcripts and ranks candidates.
Done withClaude
- What they did
- Django sends each uploaded PDF resume to Claude as a base64 document, with no parser. Claude returns a score and recommendation. For candidates scoring 60 or higher, Django triggers a Vapi outbound call. Vapi calls a custom Claude endpoint, and Claude asks four questions and ends the call with an end_call tool. A webhook then delivers the transcript for scoring.
- How it went
- Screening takes under 5 seconds per resume, and an interview takes about 5 minutes. Scores of 75 or more are shortlisted and under 50 are rejected. Scores from 50 to 75 need a human decision. The post presents it as a v1 and gives no hiring outcomes.
- Worth knowing
- The author estimates about $50/month for a small team (Claude API, Vapi calls, a small server). Voice calls need the backend exposed publicly, for example through ngrok.
Try it yourself with Claude
Build me a small script that screens resumes for the [job title] role. It should read every PDF in [folder path], score each resume 0-100 against these requirements: [list of must-haves and nice-to-haves], and write a ranked CSV with the score and a one-sentence reason for each. You are done when the CSV exists and I've reviewed the top 10; do not contact any candidate, and pause and show me the list before any step that would call or message someone.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Source: codewithmuh.com · Undated
Related use cases
Build a C compiler with a team of parallel agents
Anthropic's engineering write-up on tasking Opus 4.6 agent teams with building a C compiler.
Coach training and diet from wearable data
The poster uses Claude with Apple Health, nutrition and lifting app data, plus medication and bloodwork files, and has it maintain an HTML dashboard.
Research markets and accounts with Claude Cowork for sales
A salesperson has Claude Cowork do market and account research and prospecting via a ZoomInfo API while they take calls, manually reviewing what it pulls.
Run a solo company with eight agent departments
A founder set up CEO, CFO, COO, lawyer, accountant, marketing, CTO, and improver agents that share memory and consult each other.
Collect shower glass supplier quotes by email
OpenClaw emailed shower glass suppliers, collated responses, and handled the back-and-forth in a chat thread.
Write and review contracts with ChatGPT
The poster used ChatGPT to draft and review contracts and was impressed with the results.
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.