Case study / 03
I built Kunang as a side project during Challenge 5 at the Apple Developer Academy. It brings community mental-health intake, prioritization, scheduling and live communication into one iPhone and iPad workflow, while keeping sensitive processing and credentials in the right place.
01 / Learning & outcome
My favorite milestone was sending a real message to the WhatsApp number, watching it appear inside Kunang and continuing the conversation from the app. That closed the loop between a native interface, a serverless backend and a communication channel people already use.
Kunang pushed me beyond native iOS development into on-device AI, webhooks, databases, API security and serverless integration, all without losing sight of the community workflow the technology was meant to support.
02 / The problem
Kunang was designed for a community mental-health team in Bali. The operational challenge was not one isolated screen: it was the handoff from first contact to intake, triage, scheduling, handler assignment and continued conversation.
01 / Intake
The app can import client records from a spreadsheet instead of forcing the team to rebuild its working data manually. Structured local storage then becomes the foundation for prioritization and scheduling.
02 / Triage
Using Apple’s Foundation Models framework, Kunang performs structured triage locally, assigns priorities and supports decision-making without sending the entire workflow to a remote AI service.
03 / Scheduling
Kunang creates schedules and distributes sessions between available handlers. The goal is to transform intake signals into a practical plan while keeping the team’s workload and client urgency visible.
03 / WhatsApp relay
The part I most enjoyed was connecting Kunang to the community’s existing WhatsApp channel. I created a serverless relay with Cloudflare Workers and D1 so the mobile app could participate in conversations without carrying the WhatsApp Cloud API token.
01 / Receive
Meta forwards incoming messages to the Cloudflare Worker. The relay validates the webhook before accepting and storing the event.
02 / Guide
For new contacts, the relay asks about language, location, immediate danger and the type of support they need.
03 / Prioritize
If crisis-related language is detected, the flow immediately provides emergency information and raises the contact’s priority.
04 / Synchronize
Kunang synchronizes with the relay, displays active conversations within seconds and lets a handler reply from the app.
04 / Architecture
The mobile app and relay have different responsibilities. This boundary protects credentials and concentrates the integration rules in one server-side layer.
01 / Mobile
The native client manages the working interface, local persistence, on-device triage, schedules and the synchronized conversation view across iPhone and iPad.
02 / Relay
The serverless layer stores the access token, validates webhooks, prevents duplicate messages, keeps conversation history and manages priority queues.
03 / Platform rule
The relay centralizes WhatsApp’s messaging-window rules so the client does not need to reproduce server-side policy logic.
04 / Honest infrastructure
The architecture could be expressed with Lambda, API Gateway and DynamoDB, but the implemented version uses Cloudflare Workers and D1 on the free plan.