Prepwise

I built this project solo to combine conversational voice UX with structured interview assessment workflows, covering auth, data persistence, generation, and post-session scoring.

Highlights

• Implemented account flows with Firebase-backed signup/signin and secure session cookie handling for protected routes.
• Added authenticated app layouts that gate interview and feedback routes based on session state.
• Built interview question generation endpoint using Gemini with role, level, tech stack, and interview-type prompt inputs.
• Integrated Vapi voice sessions for live mock interviews, with event-driven transcript capture and call lifecycle handling.
• Implemented transcript-to-feedback pipeline using structured schema output for total score, category scores, strengths, and improvement areas.
• Added feedback upsert logic so users can regenerate updated evaluations for the same interview.
• Persisted interviews and feedback in Firestore collections with user linkage and created-at ordering.
• Built dashboard views for user interview history and latest interviews from other users.
• Added interview detail and feedback pages to review session context and actionable assessment output.

Overview

Prepwise is an AI-powered mock interview platform that combines question generation, live voice interview sessions, and structured feedback scoring.

The goal is to move from static practice prompts to a full interview loop with measurable output.

Product Flow

The user journey is:

• generate interview set from target role metadata
• run a live voice interview session
• capture transcript events
• produce structured feedback and score breakdowns

This gives users both practice and concrete evaluation artifacts.

Auth And Data Layer

Authentication and session management are built on Firebase. Interview and feedback records are stored in Firestore and scoped by user identity.

This keeps route protection and data access patterns straightforward.

Voice And AI Integration

I integrated Vapi for the interview call experience and Gemini for two core tasks:

• generating interview questions
• generating structured post-session feedback

Feedback generation is schema-constrained to keep output predictable for UI rendering.

Feedback Quality Surface

Each feedback record includes total score, category-level scoring, strengths, areas for improvement, and a final assessment narrative.

This allows users to revisit past sessions and compare progress over time.

Why This Project Matters

Prepwise demonstrates how I compose multiple AI capabilities into one product flow: real-time voice interaction plus structured reasoning outputs that can be persisted and productized.