AI-Powered Mock Interview & Smart CV Builder

An AI platform that simulates real interviews and generates personalized feedback and job-tailored CVs.

Comment

Inspiration

Preparing for technical interviews is challenging, especially for software engineering roles. Many candidates struggle not because they lack technical skills, but because they lack realistic practice and personalized feedback.

Most existing platforms provide generic questions and static evaluations, which do not reflect real interview scenarios or specific job requirements.

We were inspired to build a system that can simulate real, personalized interviews based on a candidate’s profile and target job, and help them continuously improve with actionable insights.

What it does

This project is an AI-powered mock interview and smart CV builder platform.

Mock Interview Simulation

Simulates real interview sessions using:

  • CV
  • Job role
  • Job description
  • Company information

Generates:

  • Technical and behavioral questions
  • Follow-up questions based on user responses

Intelligent Evaluation

Provides structured feedback across:

  • Technical skills
  • Project explanation
  • Communication
  • Behavioral performance

Each category is scored on a scale: [ Score \in [1, 10] ]

With detailed explanations and improvement suggestions.

Smart CV Builder

Builds job-specific CVs using:

  • GitHub (projects & code analysis)
  • LinkedIn (experience)
  • Interview history

How we built it

Tech Stack

Frontend

  • Next.js (App Router) – interactive UI and smooth interview experience
  • React + TypeScript – state management and type safety
  • Tailwind CSS – responsive UI with dark/light theme support
  • Zustand – lightweight global state management

Backend

  • FastAPI (Python) – high-performance API for interview sessions
  • PostgreSQL (via Supabase) – persistent storage
  • Supabase Auth & Storage – authentication and file handling
  • Redis – session memory and caching

AI System

  • LLM (Gemini / GPT-based):

    • Question generation
    • Answer evaluation
    • CV generation

  • RAG (Retrieval-Augmented Generation): Combines:

    • Job description
    • Company data
    • User CV

  • Embeddings (sBERT – Vietnamese):

    • Semantic search
    • Context retrieval

  • Vector Database (Weaviate / Qdrant):

    • Store embeddings
    • Enable semantic retrieval

Infrastructure

  • AWS S3 / Supabase Storage – file storage
  • GitHub API – project analysis
  • Environment Configuration (.env):
    • NEXT_PUBLIC_API_URL
    • SUPABASE_URL
    • GEMINI_API_KEY
    • REDIS_URL
    • MAX_INTERVIEW_QUESTIONS
    • QUESTIONS_PER_SECTION

Core Features

AI Mock Interview Engine

  • Personalized interview generation
  • Multi-step interview flow
  • Context-aware follow-up questions

Intelligent Feedback System

  • Multi-dimensional evaluation:
    • Technical
    • Communication
    • Behavioral
  • Structured scoring and explanations

Smart CV Generator

  • Job-specific CV creation
  • Avoids generic templates
  • Uses real user data

RAG-based Personalization

  • Context retrieval from multiple sources
  • Improves accuracy and relevance
  • Reduces hallucination

Conversational Interview Flow

  • Chat-based interaction
  • Maintains context using Redis
  • Dynamic question generation

Real-time Interview Experience

  • Camera and microphone integration
  • Simulates real interview conditions

Performance Tracking

  • Score tracking across sessions
  • Continuous improvement insights

Challenges we ran into

Personalized Interview Generation

Ensuring questions match:

  • Job role
  • Company
  • Candidate profile

This required combining multiple data sources effectively.

Consistent AI Evaluation

LLM outputs can be:

  • Inconsistent
  • Too generic

We addressed this with:

  • Prompt engineering
  • Structured output formats

High-quality CV Generation

Avoiding:

  • Generic CVs
  • Irrelevant content

We focused on:

  • Context-aware generation
  • Job-specific tailoring

Data Integration

Combining:

  • GitHub
  • LinkedIn
  • CV

Into a unified pipeline was non-trivial and required careful system design.

Accomplishments that we're proud of

  • Built a full end-to-end AI system, not just a chatbot
  • Successfully implemented personalized interview generation
  • Designed a multi-dimensional evaluation system
  • Created a job-specific CV generator
  • Integrated multiple data sources into a unified pipeline

What we learned

  • How to design and implement RAG-based AI systems
  • The importance of prompt engineering for structured outputs
  • Techniques to make AI outputs more consistent and reliable

  • Real-world system design challenges:

    • Scalability
    • Data flow

  • The trade-off between: [ Flexibility \leftrightarrow Control ]

What's next

  • Voice-based interviews for realism
  • Real-time coding interviews
  • Fine-tuned models for better evaluation
  • Company-specific interview datasets
  • Advanced CV analysis with deeper project insights

Built With

  • amazon-web-services
  • cache:
  • cloud:
  • db:
  • languages:-python
  • qdrant
  • redis
  • sbert
  • typescript-frontend:-next.js-backend:-fastapi-ai/ml:-llm-(gpt)
  • vector
  • weaviate
Share this project:

Updates