This is a submission for the Built with Google Gemini: Writing Challenge
What I Built with Google Gemini
I developed CV Advisor PRO, a senior career auditor designed to move beyond simple spell-checking and into high-level career coaching. Many talented professionals miss out on opportunities because they lack the "language of leadership." This tool democratizes access to the kind of feedback usually reserved for expensive executive consultants.
Google Gemini serves as the brain of the operation. Leveraging the Gemini 3 Flash Preview model via Google AI Studio, the app performs deep-tissue scans of professional documents. It uses Gemini’s massive context window and multimodal capabilities to:
- Analyze PDF structures and extract complex work histories.
- Apply the STAR method (Situation, Task, Action, Result) to transform passive duties into high-impact achievements.
- Generate an Employability Score.
Demo
Here is my video demo.
What I Learned
Building this tool taught me that career coaching is a delicate balance of data science and psychology.
- Technical Growth: I mastered prompt engineering to ensure the model maintained a "Senior Career Strategist" persona, providing "tough love" without being discouraging.
- Product Philosophy: I realized that users don't just want a better document; they want the confidence that comes from knowing their worth. I applied to my CV and it was very helpful.
- System Design: Using Cloud Run and TypeScript alongside Gemini showed me how quickly a high-speed, scalable AI microservice can be deployed.
Google Gemini Feedback
The experience with Gemini 3 Flash was a game-changer for this specific use case:
The Wins: The model's responses are consistent and useful. I applied several improvements to my own CV. Also the speed is incredible. For a real-time conversational coach, latency is the enemy, and Flash handled complex reasoning almost instantly. Its ability to "understand" the nuance between a responsibility and an achievement was far superior to previous models I've tested.
The Friction: Fine-tuning the "AI Carrer Coaching" aspect required several iterations. Initially, the model tended to be too polite (the model praised everything, even irrelevant details). Another point of friction was that it types the response much faster than it can read. That's why I limited interaction after the AI finishes speaking.
Future Support: I’d love to see even deeper integration for real-time web-scraping within the AI Studio environment to facilitate the "Live Job Comparison" feature I have planned next.
Top comments (0)