IntervU

Inspiration

Earlier this week, two of our team members had software engineer co-op interviews that didn't go as expected. They realized they could've been way more prepared for different types of questions if there was an easier way to practice. Something that didn't involve bugging friends for mock interviews or paying for expensive coaching sessions. There are a few mock interview free plans but with strict limits and other plans could be as much as $80 a month.

We wanted to build something that gives you personalized interview prep without taking up anyone else's time or breaking the bank. Just you, some AI, and the chance to practice until you're confident.

Links:

NOTE: If the backend API documentation website or the live site is slow, the backend is most likely just ramping back up from a sleeping state (hopefully)

What it does

Step 1: Upload your resume and job description and our Google Gemini AI analyzes both to understand your background and the role.
Step 2: Get personalized interview questions tailored specifically to you and the position you're applying for.
Step 3: Practice with behavioural questions and coding challenges. Write and test code directly in the platform, it can run your Python or JavaScript and validates against test cases.
Step 4: Ask for AI-powered hints when stuck, without getting the answer spoiled.
Step 5: Receive instant feedback after each question with performance insights.
Step 6: Get a comprehensive final report with your hire-ability score and actionable improvement suggestions.

It's like having a personal interview coach available 24/7, without the scheduling or cost.

How we built it

Core Stack

Next.js, React.js, TypeScript - Everything on the frontend
Go - Fast, concurrent HTTP server
MongoDB - Stores interview sessions and question banks
Google Gemini AI - Powers intelligent question customization, hints, and feedback
Go-Piston - Remote code execution for Python/JavaScript

Key Features

Smart Question Generation - Gemini customizes questions based on job title and company
Real-time Code Execution - Execute user code against test cases instantly
AI-Powered Hints - Context-aware hints that adapt to user's progress
Technical Feedback - Comprehensive evaluation with hireability scores

Backend Architecture

User Request → Go Handler → Service Layer → Repository → MongoDB
                    ↓
              Gemini AI + Go-Piston (Code Execution)

Challenges we ran into

Running out of various service's tokens while testing project (ChatGPT, ElevenLabs, etc)
Unfamiliarity with design software (Adobe Illustrator, Figma)
Integrating a code editor in the frontend
Implementing TTS and STT due to unfamiliarity which required a lot of trial and error
Figuring out how to do code execution with certain languages (ended up choosing Python & JavaScript) and how to evaluate expected output
Wanting to learn/use GraphQL but ending up making REST
Google Gemini sometimes returns JSON wrapped in markdown code blocks (``json), so had to create acleanJsonResponse()` function to call after every service call

Accomplishments that we're proud of

The overall design with the unfamiliarity and an unexpected 2 hour bus ride
Backend code evaluation and execution to be used in the technical interview part of the project
Productive use of ElevenLabs API and Google Gemini API
Using a backend language that we have little to no experience with
First time successfully deploying a backend on something (Render) other than AWS or BaaS's like Supabase
Successfully slightly altering stored behavioural questions based on info like the user's resume, job title, and seniority

What we learned

Go Backend Development: Built our first production Go API from scratch, learning concurrency patterns, HTTP handlers, and MongoDB integration
AI Integration: Successfully integrated Google Gemini for intelligent question customization, hint generation, and technical feedback
Code Execution: Implemented remote code execution using Go-Piston, handling multiple languages and test case validation
Database Design: Designed MongoDB schemas for interview sessions, question banks, and technical problems with proper relationships
Deployment: Successfully deployed a Go backend to production using Render, learning about environment variables and service configuration -App Design: Custom raccoon mascot with hand-drawn aesthetic for SFU spirit, and lighthearted feel.