PM Interview Coach
LiveAI-powered practice tool using real PM frameworks and ChatGPT evaluation.
The Problem
PM interview prep is fragmented. Candidates memorize frameworks but don't know how to apply them under pressure. Feedback comes too late to iterate.
The Approach
Build an AI coach that evaluates answers in real-time using the same frameworks interviewers use (CIRCLES, RICE, HEART, STAR). Make feedback immediate and actionable.
The Solution
An interactive tool where users answer PM questions and get structured feedback on frameworks, clarity, and strategy. They iterate until they're confident.
What
An AI-powered interview practice platform that gives you realistic PM questions and immediate, structured feedback on your responses.
You get: - Real PM interview questions across 4 categories (Product Design, Strategy, Execution, Analytical) - Instant AI evaluation on 4 dimensions: overall score, framework usage, structure, completeness - Specific strengths and gaps highlighted in every answer - Progress tracking to see improvement over time - No API key required — everything runs server-side
Why
PM interview prep has a fundamental problem: feedback lag.
You practice in a vacuum, memorize frameworks, and hope they stick. Then you hit the real interview and realize there’s a gap between understanding CIRCLES and using CIRCLES under pressure.
I built this because: - Candidates need iteration, not just memorization. Real learning happens when you answer, get feedback, adjust, and try again. - Interviewers look for reasoning, not perfection. The tool evaluates how you think, not whether you’re “right.” - The best interview prep feels like the real thing. Realistic questions, structured evaluation, honest feedback.
This tool bridges that gap. It’s the coach who gives you feedback at 2am when no actual interviewer is available.
How
The flow is simple:
- Pick a category — Product Design, Strategy, Execution, or Analytical
- Read your question — Realistic PM scenarios (e.g., “Design a feature to help users discover restaurants”)
- Answer out loud — Structure your thinking like you would in a real interview
- Get scored — AI evaluates on frameworks, structure, and completeness
- See feedback — Strengths, improvements, and suggested frameworks
- Practice again — Pick another question and iterate
Under the hood: - Questions are curated to reflect real PM interviews (not trick questions) - OpenAI ChatGPT evaluates answers against PM frameworks (CIRCLES, RICE, HEART, STAR, etc.) - Each evaluation is stored so you can track progress over time - The AI gives specific, actionable feedback (not vague praise or criticism)
Why no API key? All AI evaluation runs server-side. You don’t need to set up OpenAI or worry about costs. Just practice.
Technical Stack
- Backend: FastAPI (Python) — async, fast, built for real-time feedback
- AI Engine: OpenAI ChatGPT (gpt-4o-mini) — evaluates answers against frameworks
- Database: SQLite — stores your sessions and scores
- Frontend: HTMX + Tailwind CSS — interactive without JavaScript bloat
- Deployment: Render — single command deploy, auto-scales