DEV Community

Cover image for I Built an AI-First Portfolio Powered by Gemini
Anshul Negi
Anshul Negi

Posted on

I Built an AI-First Portfolio Powered by Gemini

New Year, New You Portfolio Challenge Submission

This is a submission for the New Year, New You Portfolio Challenge Presented by Google AI

About Me

I like building things that are simple, useful, and easy to understand.
This portfolio replaces traditional navigation with conversation, letting you explore my work and approach through an AI-powered interface.

Portfolio

Try it here → Live Demo

How I Built It

Tech stack

  • Angular
  • TailWind CSS
  • Gemini Flash Lite 2.5

Design Decisions

  • Chose a conversational portfolio instead of a multi-page layout to highlight AI-driven interaction.
  • Kept the UI simple and readable so users can focus on content without distractions.
  • Added predefined prompts to help users get started easily.

Development Process

  • Gathered all portfolio information and structured it as a single data source to enable context-based responses.
  • Initially passed the entire portfolio data with each request, which led to unnecessary API usage and performance issues.
  • Introduced AI-based intent detection to understand what the user is asking before selecting relevant data. This helped reduce context size but still required further optimization.
  • Split the portfolio data into multiple focused sections (skills, projects, experience, philosophy) to improve clarity and reduce payload size.
  • Switched to Gemini Flash Lite 2.5 for faster, more efficient intent detection and lightweight response generation.
  • Defined fixed intents for predefined prompt pills to avoid unnecessary intent detection calls.
  • Cached responses for prompts to minimize repeated API calls and improve responsiveness.
  • Added a contextual follow-up question after responses to encourage deeper exploration without overwhelming the user.

Google AI tools

  • Google AI Studio
  • Gemini

What I'm Most Proud Of

  • Used AI to rapidly build, iterate, and deploy the application, enabling fast experimentation and continuous refinement.

  • Applied intent-based AI usage with focused context and model selection to improve response quality while reducing unnecessary API calls.

  • Balanced user experience and technical implementation, ensuring the interface remains simple while the system stays efficient.

Top comments (0)