A build plan for an AI health product
Product scope, technical architecture and cost modelling for a startup in the UAE.
Move across the paths to highlight their dots. Click or tap to hold the highlight. Use the left and right arrow keys to explore, or Escape to reset. The illustration contains no customer data.
A founding team in the UAE wanted to bring personal and family health information into one AI-assisted product. Alf Viktor Williamsen (founder of Kylace) worked with the founders to define what to build, how it could work and what it could cost to operate. The advisory work produced a technical architecture summary and a path-to-market presentation.
Client recommendation
“Alf is proactive at working with the entrepreneurs to organise their business proposition, and is up-to-date with latest tools and technology. Alf pulls a build plan together which fits seamlessly into the commercial plans.”
Stuart Laird-Smith
Co-founder, stealth startup in the UAE
From product idea to system design
The advisory work addressed how family members and caregivers would access and share information. The proposed architecture covered individual records, access to dependent accounts and opt-in sharing, alongside intake from documents, connected devices, photos and voice notes.
Interface concepts showed how this could work in a native iOS app, from adding records to viewing daily summaries and preparing a report for a doctor. These concepts made the proposed product concrete enough to discuss alongside the technical design.
Connecting scope and cost
Cost scenarios compared AI models and estimated monthly operating costs at different user volumes. Scheduled AI processing, selective on-device work and per-user spending limits were considered as ways to manage those costs.
AI inference strategy
Read moreSmall tasks on device. Deeper analysis on schedule.
Proposed design
A workload follows two paths. Small tasks pass through local processing individually. Larger dots represent heavier work. They wait in a queue, then move together when the scheduled run begins. Choose a view to focus on one path. Drag horizontally or use the left and right arrow keys to pause and inspect the sequence. Use Play to resume.
AI inference strategy
The proposed strategy combines on-device processing for smaller, frequent tasks with scheduled cloud processing for more complex analysis.
Local work is bounded by device and model capability. Scheduled cloud runs can use larger models, with results arriving after the run. Choosing between them requires evaluating the task, acceptable delay and cost.
The paths illustrate placement and timing. Their dot counts, sizes and animation speed do not represent measured cost, capacity or savings. Model choices and operating costs remain subject to evaluation.
The delivery proposal set out an initial build covering accounts, data intake and the core app, followed by subscriptions, monitoring and preparation for a public launch. Production development is still ahead.