Problem
Streaming data platform that analyzes wellness signals in real time, surfaces evolving risk scores, and delivers empathetic AI coaching for patients while giving counselors a simple triage view.
PathwayFastAPIPythonStreaming DataAI CoachingCase study / interactive demo
Case study
From problem to working system
My role
AI product engineer and full-stack developer
Solution
Streaming data platform that analyzes wellness signals in real time, surfaces evolving risk scores, and delivers empathetic AI coaching for patients while giving counselors a simple triage view.
Architecture
The implementation combines the following technologies and system concerns.
PathwayFastAPIPythonStreaming DataAI Coaching
How it was built
- Built a streaming platform on Pathway and FastAPI that scores relapse risk continuously from wellness signals instead of at scheduled check-ins.
- Modelled risk from sleep, mood, craving intensity, social contact, and routine disruption, weighting recent change more heavily than absolute values.
Security decisions
- Built a streaming platform on Pathway and FastAPI that scores relapse risk continuously from wellness signals instead of at scheduled check-ins.
- Modelled risk from sleep, mood, craving intensity, social contact, and routine disruption, weighting recent change more heavily than absolute values.
- Surfaced an evolving risk trajectory so a slow slide is visible before it becomes an acute event.
- Gave counsellors a triage view ordered by risk change, so limited clinical attention goes where the trend is worst.
Major challenges
- Surfaced an evolving risk trajectory so a slow slide is visible before it becomes an acute event.
- Generated empathetic coaching messages matched to the specific signal that moved, avoiding generic encouragement that patients disengage from.
- Gave counsellors a triage view ordered by risk change, so limited clinical attention goes where the trend is worst.
Verified evidence
Results and measurable impact
- Surfaced an evolving risk trajectory so a slow slide is visible before it becomes an acute event.
- Generated empathetic coaching messages matched to the specific signal that moved, avoiding generic encouragement that patients disengage from.
Screenshots and access
Product view
Interactive Demo
A scoped, fully functional recreation of this project's core feature runs below, live in your browser. Reset it, resize it, or expand it to full screen.
relapseradar.zainkhalilkhan.com
RelapseRadar
Protective trend monitoring
RelapseRadarWorkspace2 updates
RelapseRadarCONTINUITY OF CARE MONITOR
SIGNALS LIVEPATIENT OVERVIEW · LIVE SESSION
Early signals, human response.
Transparent risk movement keeps counselors informed without turning a care relationship into a black box.
Sleep
7hrs
Mood
6/10
Cravings
2/10
HRV
62ms
Composite risk
32
Stable
0 over last 5 ticks
Risk stream · nudge signals to move it
Coach · simulated AI
Counselor triage · live patient re-ranks with the stream
#1 J.M.64
Missed two check-ins
#2 T.S.37
Cravings trending up
#3 You32
Live session
#4 A.R.21
14 days consistent
Zain Khalil Khan
Client-side sandbox. State is in memory and nothing is sent to a server.