KAIROS - AI Surf Oracle for Lisbon by Miguel Marinho

Summary

KAIROS replaces the forecast-spreadsheet ritual with a single prompt: KAI reads seven swell models across 11 Lisbon-coast spots and answers in plain English with day-and-time recommendations and a skill-aware 1-10 rating.

Problem

Checking the surf meant cross-referencing multiple swell models, wind, tide, and crowd data across many spots in spreadsheets before deciding when and where to paddle out.

My role

Miguel designed, built, and shipped the full product: the KAI conversational agent, the multi-model forecast aggregation, the rating system, and the live deployment.

Architecture

• Seven swell models are aggregated and normalized across 11 Lisbon-coast surf spots.
• KAI, a conversational AI agent, answers natural-language questions with specific day-and-time recommendations and reasoning.
• A skill-aware 1-10 rating combines condition quality (firing, solid, rideable, flat) with skill level (beginner, improver, intermediate, advanced).
• A live Carcavelos webcam feed lets surfers cross-check the AI guidance against real conditions.
• The app is deployed publicly on Vercel.

Outcomes

• Surfers get a single-prompt answer instead of manually reading forecast spreadsheets.
• Recommendations are timed and skill-aware rather than raw model output.
• The live deployment makes the project easy to inspect from the portfolio.