Interview Shadow AI

Secure login screen for candidates to continue their interview journey.
Landing page for AI-powered mock interview practice and personalised feedback.
Adaptive follow-up question asks the candidate to separate individual contribution.
Scorecard uses varied category scores based on interview response quality.
Interview setup with role, experience level, skills, and resume context.
Resume upload and highlights input to personalise interview questions.
Question 10 tests product judgment, metrics, and decision-making.
Question 1 starts with project-specific evidence and ownership.
Detailed evaluation explains evidence, improvement areas, and next actions.
Question 14 evaluates learning ability and technical growth.
Final question evaluates self-awareness, ownership, and hiring readiness.

Inspiration

Technical interviews can feel unpredictable because candidates rarely get realistic follow-up questions or useful feedback after practice. We wanted to build a space where students can practise explaining their projects, technical decisions, trade-offs, and impact before a real interview.

What it does

Interview Shadow AI is a mock technical-interview platform. Candidates can sign in, upload a resume, enter their target role and skills, complete a 15-question interview, and receive a scorecard.

The scorecard evaluates communication, technical knowledge, and confidence. It also provides strengths, areas to improve, a hiring recommendation, and detailed evidence-based next steps.

How we built it

We built the frontend with React, Vite, Tailwind CSS, Axios, and Firebase Authentication. The backend uses FastAPI, Python, SQLAlchemy, and SQLite to store interview history, answers, and scorecards.

The interview flow uses candidate role, skills, resume highlights, project context, and previous answers to guide questions. GPT-5 support is included for adaptive questions and structured feedback. We also built a demo mode so the complete 15-question flow and answer-quality scoring can be demonstrated locally without API quota.

Challenges we ran into

The main challenge was making feedback meaningful instead of giving every candidate the same high score. We improved the scoring logic to consider technical concepts, ownership, outcomes, answer clarity, and vague responses.

We also handled API configuration, local frontend-backend communication, Firebase authentication setup, database persistence, and responsive UI states.

Accomplishments that we're proud of

Built a complete 15-question technical interview experience.
Personalised questions using resume, role, skills, project context, and prior answers.
Created varied scorecards instead of fixed scores.
Added detailed evaluation with assessment, evidence, and an actionable next step for each category.
Connected authentication, interview storage, answers, and scorecards in one workflow.
Created a polished responsive UI for the full candidate journey.

What we learned

We learned how to connect a React frontend with a FastAPI backend, design structured API contracts, manage authentication, persist user interview history, and create transparent scoring logic.

We also learned that interview feedback is most useful when it explains why a score was given and tells the candidate exactly how to improve.

What's next for Interview Shadow AI

Next, we plan to add public backend deployment, richer GPT-powered follow-up questions, resume parsing, voice interview mode, company-specific interview templates, interviewer analytics, and stronger production metrics for feedback quality.

Built With

add-these-tags:-`react
ai
axios
docker
fastapi
firebase-authentication
git
github
gpt-5
javascript
machine-learning
openai-api
pdf-parsing
prompt-engineering
python
rag
render
rest-api
sqlalchemy
sqlite
tailwind-css
vector-database
vercel
vite

Submitted to

OpenAI Build Week

Created by

I contributed to developing Interview Shadow AI from idea to final demo. My work included designing the end-to-end product flow, connecting the React frontend with FastAPI APIs, improving the 15-question interview system, and refining the scoring and detailed-feedback experience.

I worked on the backend interview lifecycle, including starting interviews, saving answers, generating follow-up questions, finishing sessions, and producing scorecards. I also contributed to GPT-5 integration, prompt design, demo-mode fallback, SQLite persistence, Firebase authentication flow, responsive UI improvements, testing, debugging, GitHub documentation, screenshots, and Devpost submission preparation.

My focus was helping make the project feel like a real interview product rather than a static UI: questions use interview context, scores vary with answer quality, and feedback explains strengths, gaps, evidence, and practical next steps.

Muskankumari13 Bai
As Member 3, I handled the database, and testing for the project. I designed the SQLite database to store users, interview history, answers, and scorecards, and I also conducted end-to-end testing of the complete application (frontend, backend, and database) to ensure smooth functionality. Additionally, I prepared the README, screenshots, and documentation, and managed the GitHub repository by merging all team members' code.

Hafsa Atif
As a team member, I contributed to the frontend development of the application using React. I designed and implemented the user interface, including the landing page, login/signup, resume upload, interview, and feedback pages. I integrated Firebase authentication, created reusable UI components, and tested the frontend to ensure smooth functionality. In addition, I contributed to the project demonstration video by explaining the system features and helping present the overall workflow of the application.

Hafsa Naz Naz

Updates

Muskankumari13 Bai started this project — Jul 20, 2026 04:30 AM EDT

Leave feedback in the comments!

Log in or sign up for Devpost to join the conversation.