Contact person: AI Skill Scoring Workflow

Phone:Show

Email:Show

Location: Vallensbæk Strand, Denmark

Budget: Recommended by industry experts

Time to start: As soon as possible

Project description:
"I already have ElevenLabs streaming every conversation into my pipeline; what I’m missing is the piece that decides whether the speaker truly demonstrates the skills we’re testing for. I need an autonomous workflow that picks up each transcript, evaluates it against a predefined skills matrix, and hands my backend a clean JSON scorecard.

How the evaluation should work
The aim isn’t simple sentiment checks or keyword tallies—it’s a custom algorithm that can confirm whether the user actually lives up to the skill set claimed. You’re free to combine LLM reasoning, embeddings, or other NLP techniques, as long as the logic is transparent and defensible.

The skills to grade
Every assessment comes with four buckets I’ll supply dynamically: essential skills, essential knowledge, optional skills, and optional knowledge. The exact technical topics—programming languages, problem-solving abilities, system-design elements—are already stored in the system and passed into ElevenLabs; your workflow just needs to read them.

End-to-end flow I’m after
1. Trigger from the existing ElevenLabs webhook when a conversation ends.
2. Pull the transcript, analyse it against the supplied skills list, and generate per-skill scores plus a short rationale.
3. Return a JSON object that follows the schema I’ll share so my platform can ingest it without further parsing.

Tech comfort zone
REACT, Node.js, or similar—whatever integrates neatly with the ElevenLabs and OpenAI (or comparable) APIs. Prior experience with ElevenLabs webhooks or LLM-based grading is a bonus.

Acceptance for delivery
• Fully automated run for every conversation
• Scores 0-100 for each skill bucket and an overall pass/fail flag + plus a note for it
• JSON validates against the provided schema
• Clear, commented code and a concise README for deployment

If you can spin up this evaluator and wire it into my current stack, let’s talk details and timelines." (client-provided description)