Confetti

Framing the Problem

What is the problem you are trying to solve? Who does it affect?

My project covers the connection needs of all attendees, speakers and organisers using the power of AI. The attendees look for meaningful connections. The ones attending offline get to meet people physically to establish connections. Especially the online attendees get very little to no exposure. My project solves this problem of all attendees.

Tech conferences are massive, vibrant events, but they are incredibly overwhelming for attendees, speakers, and organizers alike.

• Attendees struggle to find the right peers or naturally break the ice in a crowded room.

• Speakers are often isolated when trying to refine their upcoming presentation pitches under tight deadlines.

• Organizers lack real-time visibility into conference engagement, making it hard to see how attendees are interacting.

This disconnected experience affects everyone trying to build a professional community or gather feedback in a high-stakes environment.

Idea Explanation

What is your idea? How does it fix the problem?

The solution is Confetti, an intelligent, all-in-one conference companion platform designed to turn conference chaos into meaningful connections. Confetti solves these pain points through four tightly integrated features:

• Network Radar: Scans nearby developers based on shared interests and instantly crafts personalized, clever, 1-sentence conversation starters.

• Talk Match: Pairs attendees and researchers together based on complementary presentation topics and overlapping technical domains.

• SpeakerPrep Engine: Acts as an automated technical reviewer, analyzing talk abstracts to deliver a score out of 100, bullet points of technical strengths, and direct feedback for improvement.

• Event Pulse: A live analytics dashboard designed for organizers and attendees to witness active platform engagement, system telemetry, and networking trends in real-time.

Implementation

How do all the pieces fit together? Does your frontend make requests to your backend? Where does your database fit in?

Confetti is engineered as a highly responsive, single-page client web application built using React and Vite, with a sophisticated UI component library layout layer for quick, smooth tab transitions across all four tools.

• Frontend-to-AI Stream: The frontend bypasses traditional server-side middleware overhead, utilizing asynchronous JavaScript fetch streams to communicate directly with official Google Gemini API endpoints (v1beta/models/gemini-2.5-flash).

• State Integration: The app uses clean, client-side reactive state arrays to process profile properties, matching calculations, and feedback loops instantly. This enables real-time calculations for Talk Match and immediate data rendering on Event Pulse without requiring heavy database queries or persistent server synchronization.

Challenges

What did you struggle with? How did you overcome it?

The biggest technical hurdle was handling real-time API rate limits and accidental quota exhaustions ("Resource Exceeded") while simultaneously testing high-volume generation calls for multiple features. I overcame this by analyzing browser console error responses, refactoring internal catch blocks to gracefully handle network dropouts, and shifting the backend infrastructure to an isolated, fresh client project environment. This successfully unlocked full performance for all concurrent user tools.

Accomplishments

What did you learn? What did you accomplish?

I successfully developed and shipped an interactive, four-feature ecosystem that contextually evaluates engineering data on the fly. I mastered how to structure prompts for drastically different outputs (from short icebreakers to long-form technical evaluations), handle loading and error states across multiple concurrent async workflows, and build an interface that serves attendees, speakers, and organizers at the exact same time.

Next Steps

What are the next steps for your project? How can you improve it?

The immediate next step for Confetti is to implement a central global state context layer to seamlessly bridge the feature tabs, allowing an interaction on Network Radar or SpeakerPrep to automatically broadcast a live tracking card directly onto the Event Pulse dashboard. Long-term goals include implementing a Node.js/Express backend paired with a PostgreSQL or MongoDB database to securely store attendee profiles, save generated speaker critiques, and implement live WebSockets for instant user-to-user chat routing.