AI PrepPulse is the ultimate AI-powered career accelerator. It combines technical mock interviews, resume analysis, and portfolio reviews into a single, comprehensive readiness assessment. Powered by Google Gemini 1.5 Flash, it delivers instant, personalized feedback to help you land your dream job.
Finding a job in tech is harder than ever. Candidates struggle not just with coding, but with communicating their value and optimizing their resume for modern ATS systems.
AI PrepPulse solves this by aggregating 4 critical pillars of interview readiness into a single, 10-minute assessment:
- Technical Proficiency: Deep-dive coding and system design questions.
- Resume Impact: Structural and keyword analysis of the user's uploaded resume.
- Communication: Evaluation of clarity, structure, and conciseness.
- Portfolio Quality: Assessment of project descriptions and demonstrable work.
Why it's different: Unlike generic coding platforms, AI PrepPulse acts as a holistic career coach, using Google's Gemini 1.5 Flash to generate a startup-specific readiness score and a detailed, prioritized roadmap for improvement.
graph TD
A[User] -->|Starts Assessment| B(Frontend: React/Vite)
B -->|Uploads Resume| C{Resume Parsing}
B -->|Submits Answers| D{AI Analysis Engine}
C -->|Extracted Text| D
D -->|Prompt Engineering| E[Google Gemini 1.5 Flash]
E -->|Structured JSON Response| D
D -->|Scoring & Feedback| F[Results Dashboard]
F -->|Generates PDF| G[Downloadable Report]
We use a sophisticated, multi-stage prompting strategy to analyze the candidate's Resume, Technical Answers, and Behavioral responses in one go. This ensures the feedback is context-aware—identifying gaps between what a candidate says they know and what their resume shows.
Gone are the complex, developer-centric metrics. Our dashboard is designed for clarity and motivation:
- Readiness Score: A clear 0-100 metric with status levels (Elite, Solid, Growth).
- Visual Breakdowns: Glassmorphism-styled charts and progress rings.
- Plain English Explanations: "What does this score mean?" sections for accessibility.
Users can download a professional, print-ready PDF report of their assessment.
- Theory & Methodology: Explains how the score was calculated.
- Executive Summary: High-level insights for quick reading.
- Visuals: Simplified charts optimized for print (A4 format).
- Powered by
html2pdf.jswith a custom, hidden rendering engine.
| Component | Technology |
|---|---|
| Frontend | React (Vite), Tailwind CSS |
| AI Engine | Google Gemini API (gemini-1.5-flash) |
| State Management | React Context API |
| PDF Generation | html2pdf.js |
| Parsing | pdfjs-dist (Resume Parsing) |
- Modular Architecture: Separated concerns for API Services, Context, and UI Components.
- Robust Error Handling: "Graceful degradation" logic ensures the app provides value even if the AI service experiences partial outages.
- Responsive Design: Fully optimized for Desktop, Tablet, and Mobile.
-
Clone the repository
git clone https://github.com/yourusername/ai-preppulse.git cd ai-preppulse -
Install dependencies
npm install
-
Set up Environment Variables Create a
.envfile in the root directory and add your Google Gemini API key:VITE_GEMINI_API_KEY=your_api_key_here
-
Run the development server
npm run dev
-
Open in Browser Navigate to
http://localhost:5173(or the port shown in your terminal).
Short Answer: We prioritized Application Intelligence over Model Training. Detailed Answer:
- The Problem is "Subjective Critique," not "Classification": Training a model from scratch to understand nuances in resume quality and coding style would require massive datasets and months of compute.
- Leveraging Foundation Models: Google's Gemini 1.5 Flash is already trained on billions of lines of code. Instead of reinventing the wheel, we use advanced Prompt Engineering to direct this massive intelligence, allowing us to build a feature-rich product in a single hackathon.
- Architecture Efficiency: By bypassing a heavy Python backend for inference, our app is Serverless, faster, and easier to scale.
We use a multi-shot prompting strategy with strict JSON schema enforcement. We don't just ask "Is this good?"; we provide the AI with a rubric and specific criteria for "Elite," "Solid," and "Growth" tiers. This grounds the AI's hallucinations and ensures consistent, actionable feedback.
Yes. The parsing happens in-memory on the client (using pdfjs-dist) and is sent to the Gemini API only for analysis. We do not store your resume on any persistent database. Once the session ends, the data is wiped. This "privacy-by-design" approach makes the tool safe for quick demos.
The score is a weighted average derived from the AI's analysis of 4 pillars:
- Technical Accuracy (40%): Correctness and optimality of code.
- Communication (20%): Clarity and structure of explanations.
- Resume Impact (25%): ATS compatibility and verifiable metrics.
- Portfolio (15%): Project complexity and documentation.
- Voice Mode: Integrating Speech-to-Text for real-time verbal mock interviews.
- Browser Extension: Overlaying specific feedback directly on LeetCode/HackerRank problems.
- Company Specifics: Customizing the assessment rubric for specific companies (e.g., "Googleyness" or Amazon Leadership Principles).
- Ankit - Full Stack Developer & AI Engineer
Built with ❤️ for the Hackathon 2026.