Constraint / 01
whyThe friction
The design question was how a wellbeing interface could stay calm, preserve privacy and surface trusted crisis resources without relying on a generative response at the most sensitive moment.
Project index
Case file 03 / 05
AIWellbeing prototypeAccessibilityPrivacy
Silent Help
- Discipline
- Web App
- Context
- Personal
- Role
- Product developer and designer
- Timeline
- Sep 2024 - Present
- Duration
- Ongoing
- State
- Live · in development
01<1.5s
SOS load · design target
02<60s
To calming exercise · target
03AES-256
Field Encryption
04Privacy
GDPR-informed design
01 / The brief
The reason this needed to exist.
Silent Help is a personal, non-clinical prototype exploring privacy-first wellbeing support. A three-tier pathway model (HIGH, MID, LOW) routes the interface to different depths of support. The prototype includes AES-256-GCM field encryption, an offline-first SOS pathway with a sub-1.5-second design target, and deterministic flows that avoid generative AI in high-risk scenarios. It is not clinically validated and is not medical advice.
Response / 02
howThe intervention
The prototype explores pathway intelligence across HIGH (crisis), MID (bridge), and LOW (maintenance) support tiers. Its SOS pathway is offline-first and pre-rendered toward a sub-1.5-second load target. High-risk flows are deterministic, while AES-256-GCM field encryption and PII scrubbing are used as privacy-design exercises before any AI call.
02 / System logic
Architecture before ornament.
System backbone / documented
Architecture file 03
Next.js App Router with three-tier pathway engine (HIGH/MID/LOW), offline-first SOS rendering, AES-256-GCM field encryption, and dual-gate crisis detection
Next.jsTypeScriptTailwind CSSFramer MotionClaude APIAES-256-GCM
Constraint log
Solved through implementation
- 01
Building an offline-first SOS pathway with a target load of under 1.5 seconds using pre-rendered crisis resources
Resolved - 02
Prototyping dual-gate risk detection that swaps in deterministic crisis-resource cards
Resolved - 03
Designing a three-tier pathway engine (HIGH/MID/LOW) that routes users to the right support depth
Resolved - 04
Achieving AES-256-GCM field-level encryption with PII scrubbing before any AI call
Resolved - 05
Creating a calming UI designed to make a breathing or grounding exercise reachable within 60 seconds
Resolved
03 / Execution
Capabilities with a job to do.
Pathway Intelligence
A living three-tier engine routes you to HIGH, MID, or LOW support — the right depth for the right moment
Privacy by Architecture
AES-256-GCM field encryption, PII scrubbing before any AI call, and strict data sovereignty by default
Crisis-Aware Safety Net
Dual-gate detection swaps in deterministic resource cards when risk language is found. This is a designed prototype mechanism, not a clinically validated system
Calming Exercise Path
Breathing exercises, body scan and 5-4-3-2-1 grounding, pre-rendered toward a sub-60-second access target
04 / Ownership
What I personally moved forward.
Role on this build
Product developer and designer
Scope spans product judgement, technical execution, interface craft, and release detail where the project required it.
- 01
Architected the full-stack Next.js application with pathway intelligence routing engine
- 02
Designed and implemented the three-tier support system (HIGH/MID/LOW) with deterministic crisis flows
- 03
Built offline-first SOS pathway targeting a <1.5s load with pre-rendered crisis resources
- 04
Implemented AES-256-GCM field-level encryption with PII scrubbing pipeline
- 05
Prototyped dual-gate risk detection with deterministic crisis-resource cards
- 06
Designed calming UI with Framer Motion, targeting a calming exercise reachable within 60 seconds
- 07
Integrated semantic journalling with encrypted storage
- 08
Applied GDPR-informed privacy-by-design architecture
05 / Reflection
The work changed the next decision.
What entered the toolkit
- 1
Privacy-by-architecture patterns with field-level encryption and data sovereignty
- 2
Offline-first pre-rendering strategies for crisis-critical applications
- 3
Deterministic vs generative AI: knowing when NOT to use AI in high-risk scenarios
- 4
Designing empathetic interfaces that convey calm through motion and typography
- 5
GDPR-informed, privacy-first data handling in mental health contexts
The polished surface is only the final layer. The decisions underneath are the real project.
Continue the evidence
More case files.
05 projects / one practice
