About CareerPilot

Inspiration

It all started with my own quest to get a job. I quickly discovered the frustrating reality of the modern job market: every single application needs a unique CV. Every interview requires different preparation. And to truly excel, you need to master a new set of tools for each specific role.

I realized that "getting the job" was only half the battle; the other half was figuring out how to excel at it. I was overwhelmed by the need to be an expert on day one. This inspired me to build CareerPilot. I didn't just want a "job-finder"; I wanted to build an "AI Co-Pilot" that would give me the entire playbook—from the tailored CV to the interview scripts, and most importantly, the exact AI tools and workflows to be a top performer from the moment I was hired.

What it does

CareerPilot is a "One-Click AI Career Playbook Generator."

It's a polished Progressive Web App (PWA) where a user uploads their CV (PDF) and pastes in the text of any job description. In seconds, the app generates a complete, personalized, multi-part "Career Playbook" which is rendered directly in the app and downloadable as a PDF.

This playbook includes:

• A Job-Winning CV Rewrite: A full rewrite of their CV, tailored by AI for that specific job.
• Detailed Interview Scripts: Likely technical and behavioral questions, with strong, STAR-method answers crafted from the user's experience.
• An 'AI Performance' Plan: A secret weapon that provides a custom guide on using the latest Google AI tools (like Gemini and NotebookLM) to solve the job's key problems.

How we built it

As a non-developer, my mission was to "vibe-code" this entire project. My strategy was to act as the "Director" and use Google's AI tools as my "engineering team."

The project is built as a fully decoupled, serverless application on Google Cloud:

• The Backend "Engine": I used Firebase Studio's AI-powered IDE to generate the entire backend. This is a sophisticated Python "Orchestrator" API (using FastAPI) that I deployed as a containerized service on Google Cloud Run. This service handles all the heavy lifting: extracting text from PDFs, saving data to Firestore, and making all the calls to the Gemini API.
• The Frontend "Dashboard": I used Google AI Studio to generate the entire frontend. This is a polished, installable PWA (HTML, Tailwind CSS, JavaScript) that's also deployed as its own separate service on Google Cloud Run.
• The "AI Brain": The app's magic comes from two different Gemini API calls. It uses the standard Gemini API for the CV rewrite and interview prep. But for the "AI Performance Playbook," it uses Gemini with Google Search grounding to ensure its recommendations are always up-to-date.

Challenges we ran into

My biggest challenge was that "vibe-coding" is not magic. My first attempts were a struggle. I gave high-level prompts like "build a backend" to AI tools, and they struggled, failed, or produced code that was impossible to deploy. I was trying to build a race car engine in a bicycle shop, and I quickly realized that a high-level idea is not enough to get a high-quality, deployable product from an AI.

Accomplishments that we're proud of

Our greatest accomplishment is building a truly professional, two-service cloud application without being a traditional developer. We pivoted from a simple "download" button to a fully polished, multi-page PWA experience, complete with a landing page, animated loading screens, and in-app results.

We're incredibly proud that we overcame the "vibe-coding" challenge by creating a detailed, 13-task engineering plan. This "blueprint" allowed us to direct the AI tools with precision, building a robust, scalable application that meets all the hackathon's core and bonus requirements.

What we learned

I learned that to get a great output, you must provide a flawless input. The true power of AI-assisted development isn't about replacing the developer; it's about empowering the "director" with a clear, well-architected vision.

Instead of asking the AI to "build a backend," we gave it precise prompts like, "Write one function to extract PDF text." The AI excelled at these small, specific tasks. This granular plan was the key to success.

What's next for CareerPilot

This is just the beginning. The "One-Click Playbook" is the core, but the platform is built to expand. Next steps include:

• AI Career Coach: Integrating a voice- and chat-based AI companion to answer follow-up questions about the playbook.
• Browser Extension: A Chrome extension to analyze job descriptions "in-place" on sites like LinkedIn.
• Saved Playbook Library: A full user account system for saving, comparing, and managing all generated playbooks.
• Automated Skill-Growth Plan: Connecting the "AI Performance Playbook" to learning resources (like Google Career Certificates, Youtube) to create a trackable upskilling plan.

Built With

• css
• docker-(to-containerize-both-services)-cloud-services:-google-cloud-storage-(for-pdf-cv-uploads)-database:-google-firestore-(to-store-job/cv-text-and-the-final-playbook)-apis:-google-gemini-api-(for-cv-rewrite-and-interview-prep)
• for
• google-gemini-api-with-search-grounding-(for-the-ai-performance-playbook)-other-technologies:-html
• javascript-(for-the-frontend)-frameworks:-fastapi-(python-backend)
• jspdf-(for-pdf-generation)
• languages:-python-(for-the-backend)
• markdown
• progressive-web-app-(pwa)
• showdown.js
• tailwind-css-(frontend-styling)-platforms:-google-cloud-run-(to-host-both-frontend-and-backend-services)