DEV Community

Cover image for Geometric Journal: AI Portfolio powered by Gemini 1.5
Samuel Chimmy-Jadeofwallstreet
Samuel Chimmy-Jadeofwallstreet

Posted on • Edited on

Geometric Journal: AI Portfolio powered by Gemini 1.5

New Year, New You Portfolio Challenge Submission

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


Overview

This portfolio replaces the static resume with an AI Executive Assistant that can explain my work, answer questions contextually, check my real-time availability, and book meetings autonomously.

Deployed on Google Cloud Run and powered by Google Gemini, the goal was to explore what happens when a portfolio behaves like a product—interactive, adaptive, and intelligent—rather than a passive document.


About Me

I’m Samuel, a full-stack Web3 developer with over five years of experience architecting decentralized systems and high-performance web applications.

My core stack includes React, TypeScript, and Solidity, with a strong emphasis on performance, inclusive design, and expressive interfaces.

This portfolio is my attempt to move beyond the traditional résumé and build what I call a Geometric Journal—a space that feels as structured as code, but as warm and human as a handwritten note.


Live Portfolio

--labels dev-tutorial=devnewyear2026


Why This Stands Out

  • Innovation & Creativity: The portfolio acts as an autonomous AI agent, not a static website or simple chatbot.
  • Technical Implementation: Real Gemini tool-calling, Cloud Run deployment, and Google Apps Script middleware power live scheduling and contextual responses.
  • User Experience: Conversational navigation, procedural animation, and a living interface reduce friction and invite exploration.

Design Philosophy — Hand-Drawn Logic

The interface is built around a “Hand-Drawn Logic” system: strict geometric bento grids combined with organic typography and tactile interaction.

This creates a visual language that mirrors how developers think—structured, but expressive.


The Soul: Procedural Animation

A cinematic HTML5 Canvas overlay introduces two digital entities:

  • The Architect — a cynical backend developer
  • The Optimist — an enthusiastic frontend developer

They treat the DOM as a physical space, jumping between cards using gravity-based parabolic arcs, tracking the user’s cursor, and breaking the fourth wall to comment on the interface and code quality.

This layer exists to add personality and memorability without compromising performance.


Tech Stack

Frontend

  • React 19 + TypeScript
  • Tailwind CSS
  • Framer Motion + HTML5 Canvas

Backend & Infrastructure

  • Node.js on Google Cloud Run
  • Google Apps Script (Calendar API middleware)

AI Engine

  • Google Gemini 1.5 Flash, configured via Google AI Studio using the @google/genai SDK

🧠 AI Executive Assistant (Powered by Gemini)

The core differentiator of this portfolio is the AI Executive Assistant (AIChat.tsx).

Using Google AI Studio, Gemini 1.5 Flash is configured to act as my personal liaison, with full awareness of my projects, bio, and availability.

⚡ Why Gemini 3.0 Flash Preview

Ultra-Low Latency (Flash Advantage)

For conversational UX, time-to-first-token matters most. Gemini 3.0 Flash responds almost instantly, keeping the interface fast, fluid, and human critical for a real-time portfolio assistant.

Reliable Agent & Tool Calling

This assistant is an agent that decides when to call tools like check_calendar or create_booking. Gemini 3.0 preview significantly improves function calling accuracy over 1.5, strictly adhering to JSON schemas and reducing hallucinations.

Cost-Efficient for Public Deployment

As a public portfolio, every visitor interaction matters. Flash models offer Pro-level tool intelligence at a fraction of the cost, enabling high traffic without API throttling or runaway bills.

SDK-Aligned Choice

Per Google GenAI SDK guidance, gemini-3-flash-preview is the recommended model for real-time, interactive, user facing experiences, reserving Pro models for heavy reasoning tasks.

Gemini 3.0 Flash delivers agent-grade intelligence with instant responses and scalable economics the optimal choice for this system.


📅 Real-Time Availability

A custom scheduling interface allows visitors to book meetings directly from the portfolio.

  • Frontend: Custom calendar UI with timezone awareness
  • Backend: Google Apps Script acting as a secure proxy to the Google Calendar API
  • Security: No API keys are exposed client-side

🧪 Trade-offs & Constraints

  • Google Apps Script was chosen over a full backend to reduce infrastructure overhead and improve security.
  • The AI assistant is intentionally scoped to portfolio-related tasks to avoid feature bloat.

🏆 What I’m Most Proud Of

I’m most proud of how quickly the AI Executive Assistant moved from concept to production.

Using gemini-3-flash-preview and function calling, a fully autonomous assistant was built in under 30 minutes, transforming the portfolio from a passive document into an active digital employee.

Watching the AI transition from chat to rendering a custom success card after booking a real meeting is a clear example of how AI can enhance user experience without becoming a gimmick.


VIDEO DEMO

SOURCE CODE

Top comments (0)