CoastCast

Overview

CoastCast is a full-stack beach conditions app built for Michigan beachgoers. It surfaces real-time weather, water, and crowd data so users can make informed decisions about their beach visits — all powered by live API integrations and on-device machine learning.

The Problem

Michigan has some of the best freshwater beaches in the world but finding reliable, real-time conditions data in one place is nearly impossible. Beachgoers have to check multiple sources for weather, water quality, and crowd levels — or just show up and hope for the best.

The Solution

We built CoastCast to consolidate all of that into a single native iOS experience. The app pulls live data from multiple APIs, displays beach locations on an interactive map, and uses an on-device ML model to predict how crowded a beach will be — without sending any data to a server.

Our Role

CH Studios owned end-to-end development of CoastCast. This included building the Python FastAPI backend, integrating four external APIs, designing and training an XGBoost crowd prediction model, converting it to CoreML via ONNX for on-device inference, and building the full SwiftUI frontend with live map annotations and real-time data sync.

Tech Stack

• SwiftUI — native iOS frontend with live map annotations
• Swift Concurrency — async data fetching and real-time sync
• FastAPI — Python backend integrating NWS, NDBC, TomTom, and holiday APIs
• CoreML — on-device crowd prediction model
• XGBoost — trained crowd prediction model converted via ONNX

Outcome

CoastCast is currently live on TestFlight and in active development. On-device inference eliminated backend inference costs entirely. Iterating based on user feedback with a full App Store release planned.