AI Mock Interview SaaS Starter Kit – Django Voice AI App + Full Source Code

This is the complete source code behind a live, publicly launched AI mock interview product (originally shipped on Product Hunt and actively used by real candidates). Instead of building a voice-based AI interview app from zero — resume parsing, speech capture, AI-driven question generation, scoring — you get the entire working codebase, prompts, and build notes to launch your own version or adapt it into a new product.

Try the live product this codebase powers: ai-interviewer.tech — upload a resume and run a real mock interview, free, no account needed.

See It In Action

Want to try it yourself before buying the source code? Run a free mock interview on the live app — no signup required.

The Real Cost of Building an AI Interview Tool From Scratch

A working AI interview app isn't just "call an LLM API." It requires resume parsing across PDF and DOCX formats, reliable voice capture in the browser, an AI agent workflow that resists manipulation, and a deployment setup that survives real traffic. Most solo builders lose weeks to edge cases — broken transcripts, upload vulnerabilities, prompt injection attempts, migration conflicts — before ever shipping. This starter kit skips that phase entirely.

What's Included In The Starter Kit

• Full Django backend source code (models, views, management commands), built with a clean, modular architecture designed to scale
• Resume parsing for PDF and DOCX (including scanned-PDF detection)
• AI-driven interview question generation, built on the OpenAI Agents SDK, with tuned prompt templates included
• Voice-based answer capture flow (browser speech recognition + audio recording)
• Whisper API fallback transcription for failed browser speech capture
• AI scoring and feedback report generation logic
• Prompt injection defenses tested against real adversarial input, including multi-language injection attempts
• Device/IP-based rate limiting (no login required, ready as-is or extendable to accounts)
• Deployment notes: Gunicorn + Nginx + SSL/Certbot configuration used in production
• Written build notes documenting real bugs fixed during development

Tech Stack & Architecture

• Backend: Django, Python — organized as a clean, professionally structured workflow rather than a single-script prototype, built to be extended and scaled
• AI layer: OpenAI Agents SDK orchestrating question generation and scoring, with fallback transcription through OpenAI's Whisper API
• Resume parsing: python-docx and pdfplumber, with scanned-document detection
• Voice capture: browser MediaRecorder + Speech Recognition APIs
• Deployment: Gunicorn, Nginx, Let's Encrypt/Certbot, tested on a Linux VPS
• Data cleanup: cron-based soft-delete management command for expired interview data

Security & Guardrails

This codebase reflects fixes and defenses put in place during real production use, not a theoretical checklist:

• Prompt injection defenses: the AI agent is built and tested to ignore embedded instructions coming from user input — including attempts made in languages other than English — rather than following them
• Path traversal protection on file-serving views (safe path resolution instead of raw path joins), preventing arbitrary file access
• Magic-byte file validation on uploads — files are checked by their actual binary signature (not just file extension) before being accepted as PDF or DOCX
• Rate limiting to prevent abuse of AI-powered endpoints
• Mass-assignment protection on user-submitted feedback data
• CORS and SSL configured for production use

Note: this reflects fixes and defenses applied through real-world development, adversarial testing, and live usage. It is not a substitute for an independent security audit before you launch your own version.

Pros & Cons

Pros: built from a live, real-world product — not a tutorial project; full AI agent pipeline already working end-to-end on the OpenAI Agents SDK; clean, professionally structured codebase built to scale, not a throwaway prototype; tested prompt injection defenses instead of a naive AI wrapper; deployment-tested, not just "works on my machine."

Cons: no built-in payment/billing system (the original product is free, no-login by design); you'll need your own OpenAI/Whisper API keys and will bear usage costs; some Django and Linux deployment familiarity is expected to self-host.

Starter Kit vs. Building From Scratch

Building this pipeline from zero commonly takes several weeks for a solo developer once resume parsing, voice capture edge cases, AI agent workflow design, and prompt injection hardening are accounted for. This starter kit compresses that into a working, tested foundation you can launch or customize in days.

Who This Is For

Solo developers and small teams who want to launch an AI-powered interview, coaching, or assessment product without starting from a blank repo — or who want a real, working reference for building secure, scalable AI agent applications with Django and the OpenAI Agents SDK.