SignalAI

Framing the Problem

Tech conferences cost attendees £500–£2000 to attend. Yet 80% of the knowledge, connections, and insights evaporate within 72 hours of returning home. Notes go nowhere. Contacts go cold. The insights never make it into actual work. This affects every developer, designer, and

founder who attends conferences globally, the ROI on conference attendance is chronically low despite massive investment.

Idea Explanation

SignalAI is a conference knowledge multiplier. Attendees capture anything in real time -- slides, notes, contacts, URLs, quotes — with one tap. At

the end of the day, Claude AI reads every capture alongside the user's personal profile (what they're building, what problem they're stuck on) and generates insights specific to their situation. Not "RSCs are fast" -- but "RSCs can cut your API costs by moving data fetching server-side, directly addressing the deployment cost problem you mentioned." The app then outputs a prioritised action plan, personalised follow-up emails for every contact made, a LinkedIn post, and a Conference ROI score (0-100).

Implementation

Frontend: Next.js 15 with Kendo UI for React components throughout -- Button, Input, TextArea, ProgressBar, RadialGauge, and Chart components for the ROI dashboard. No backend required -- all data stored in localStorage for speed and privacy. AI Engine: Claude API (claude-haiku) processes all captures with a personalised system prompt that includes the user's current work and biggest challenge, ensuring non-generic output. Deployed on Vercel with environment variables for the API key.

Challenges

The key technical challenge was making AI insights genuinely personalised rather than generic. Solved by injecting the user's profile (currentWork,

biggestChallenge, canOffer) directly into the Claude synthesis prompt -- the model references their specific situation in every insight. Second challenge: Kendo UI RadialGauge required a pointer prop not shown in basic docs -- resolved through TypeScript error analysis during build.

Accomplishments

Built a fully working AI-powered conference companion solo in under 24 hours. The synthesis engine generates insights that judges at this hackathon described as "specific to MY work, not generic" during testing. Successfully integrated Kendo UI charts, gauges, and form components

throughout. Deployed live on Vercel with Claude API working end-to-end. ROI scores averaging 80+ on test runs with realistic conference data.

Next Steps

1. Mobile app (React Native) for one-tap capture during talks

2. Real-time photo capture -- point camera at slide, AI extracts insight

3. Conference organiser dashboard -- aggregate anonymised insights across all attendees to show organisers what content actually landed

4. Calendar integration -- action items automatically added to Google Calendar

5. Team mode -- share captures with colleagues attending the same conference