AI Interview Dashboard (Fullstack)

What I created

• Frontend: React + Vite + Tailwind (in frontend/)

  • Candidate interface supports video recording with single-answer enforcement, live preview, and optional resume upload for role-specific questions.
  • Recruiter dashboard allows interview creation, candidate list, rubric editing, and reviewing AI-generated summaries and evaluations.
  • Rubrics are stored in browser localStorage, can be extended to backend or Sanity/GROQ.

• Backend: FastAPI (in backend/)

  • Stores data in local JSON files (mocked GROQ-style storage)
  • Handles candidate uploads, converts videos to audio, transcribes answers using Whisper or AssemblyAI, and generates LLM-based skill evaluation.
  • Supports AI-generated role-specific introductions and dynamic question generation based on role description and resume.

• Interview flow:

  • Candidates see one question at a time, record and submit answers sequentially.
  • Recruiters can upload resumes, create interviews, and view structured AI-based skill evaluations and summaries.

---

How to run (production-like on your machine)

1. Backend

1. Create a Python virtual environment and install dependencies:

   python -m venv .venv
   source .venv/bin/activate
   pip install -r backend/requirements.txt

2. Run backend:

   uvicorn backend.main:app --reload --port 8000

3. Backend endpoints will be served on http://localhost:8000

2. Frontend

1. Install Node dependencies:

   cd frontend
   npm install
   npm run dev

2. By default, Vite serves on http://localhost:5173 (or 3000)

   • Frontend expects backend endpoints under /api/*.
   • Use the Vite proxy in vite.config.js or a reverse proxy to connect frontend to backend.

3. Notes

• Replace AssemblyAI or other integrations in candidate component and backend upload handling as needed.
• Rubrics are stored in browser localStorage but can be migrated to backend storage or Sanity with GROQ credentials.
• Candidate flow enforces single-answer per question and sequential question submission.
• AI-powered features include role-specific intro, dynamic questions, transcripts, summaries, and LLM-based evaluation.

---

AssemblyAI integration

For server-side transcription using AssemblyAI, set your API key as an environment variable:

export ASSEMBLYAI_API_KEY=your_key_here

• Backend will automatically upload submitted video/audio files to AssemblyAI.
• Transcripts will be updated in backend/data/interviews.json under the relevant interview object once ready.

---

Vite Proxy

The project includes vite.config.js configured to proxy /api to http://localhost:8000.

• Frontend dev server runs on port 3000.
• This setup allows frontend API calls to work seamlessly during local development.

---

Environment Variables (.env)

Create a .env file at the project root or set the following environment variables:

GROQ_API_KEY=...
SANITY_PROJECT_ID=your_project_id
SANITY_DATASET=production
ASSEMBLYAI_API_KEY=...

---

Key Features Recap

• AI-powered role-specific introduction for candidates.
• Dynamic interview question generation based on role description and optional candidate resume.
• Sequential question flow: candidates see one question at a time.
• Browser-based video recording with live preview.
• Resume upload for context-aware question generation.
• AI transcription using Whisper or AssemblyAI.
• LLM-based summary and skill evaluation for recruiters.
• Recruiter dashboard with rubrics, candidate list, interview links, and evaluation reports.
• Modular and extensible architecture ready for cloud storage, databases, and advanced AI features.

---

Quick Start Example

1. Candidate receives interview link.
2. System shows a role-specific introduction.
3. Candidate optionally uploads resume.
4. Candidate sees first question, records answer, submits.
5. Next question appears, repeats until all questions are answered.
6. Backend generates transcripts, AI-based skill evaluation, and stores results.
7. Recruiter reviews structured reports and decides next steps.