OCB
OnCall Balance
Your personal AI wellbeing companion for engineers.
Neuroscience & Mental Health
Rootly Sub-Challenge
CBC McGill 2026
Jeboy Compuesto · Séphorah Aniambosou · David Li
Private · Local · Powerful
THE PROBLEM
Engineers Are Burning Out.
2 AM
On-call pages
wake engineers at night
Sleep disrupted. Recovery never comes.
14%
Engineers report higher
stress post-incident
Support drops from 55% → 44% after close.
0%
Burnout risk visible
in engineering tooling
It never shows up in a Jira ticket.
"On-call rotations are a hidden health risk. Burnout builds in silence and never shows up in the metrics."
OUR PROCESS
Research-Driven from Day One.
Before writing a line of code, we grounded the product in real SRE data and validated the UX through FigJam ideation and Figma prototyping.
01
Research
SRE Report 2025 · Catchpoint (n=301)
02
Ideation
FigJam · team of 3
03
Prototyping
Figma · Stitch
Key insight: "Incidents don't end when they're over." The post-incident stress gap is the human cost OnCall Balance addresses, validated by data before we wrote a single line of Flutter.
OUR SOLUTION
Privacy-First Wellbeing for Engineers.
An AI assistant that combines Apple Watch health signals with Rootly API work data, processed entirely on your device to detect burnout risk early and suggest actionable resources.
What makes it different
All data stays on-device.
Zero employer access. Fully opt-in.
01
Health Data
Sleep duration and fragmentation count from Apple Watch via HealthKit.
02
Work Data
On-call shifts, incidents, and after-hours pages via Rootly API.
03
AI Analysis
Correlates signals locally to surface stress patterns and burnout risk.
04
Resources
Mindfulness, breathing, peer mentors — suggested, never forced.
We are not diagnosing burnout. We are detecting risk patterns and suggesting preventative actions.
THE PRODUCT
WHO WE'RE BUILDING FOR
Real Engineers. Real Problems.
A
Alex
Senior Backend Engineer
On-call 2 weeks per month
Averages 5.5h sleep during incidents
Multiple high-severity incidents per week
No visibility into his own burnout risk
Won't accept employer tracking his health
SEC
Privacy
Health and work data stays on-device. Never accessible by the employer.
SIG
Early Awareness Signal
Know when burnout is building before it becomes a crisis.
AID
Actionable Help
Real resources and peer support, not just dashboards.
HOW IT WORKS
Data Flow & Architecture.
SRC
01
Data Sources
Apple Watch / HealthKit
Rootly API
→
SEC
02
Local Data Layer
Correlation runs locally
No external transmission
→
AI
03
AI Analysis
Baseline comparison
Pattern detection (sleep ↓ + workload ↑)
Rule-based risk scoring
LLM generates natural language insight
→
OUT
04
Insight + Action
Dashboard, alerts, resources
Optional peer support
Privacy Guarantee. All computation happens on-device. Zero employer visibility. Everything is opt-in.
TECHNICAL STACK
Architecture & Implementation.
UI
Frontend
Built in Flutter, deployable on both iOS and Android (iOS only)
Onboarding: users see exactly what data is read before anything runs
Mock fallback ensures the demo never crashes if a live API fails
API
Data Integration
HealthKit: sleep data stored natively on-device by Apple
Rootly API: incidents, after-hours pages, on-call shifts
No custom backend server: correlation logic runs on the device
AI
AI Engine
StressCorrelator: explicit weighted thresholds compute risk level
Claude API (claude-sonnet-4-6) generates recommendation text only
AI communicates the result, the engineer owns the decision
SEC
Privacy & Security
No custom database, HealthKit handles on-device health storage
Correlation thresholds are named constants, fully explainable
PIPEDA / PHIPA by design, HIPAA-ready for US scaling
PRODUCT FEATURES
Four Core Screens.
DASH
01
Dashboard
Health metrics, today's workload, and AI-generated daily insight at a glance.
Sleep duration & fragmentation count
Incident severity breakdown
On-call status + risk level
TRND
02
Trends
Pattern recognition across work and health signals over time.
On-call spike detection
Mid-week drift patterns
Context switching cost analysis
RES
03
Resources
Curated, evidence-based wellness resources, locally suggested.
Mindfulness & breathing exercises
Stress-relief micro-activities
Confidential counseling links
SUPP
04
Support
Connect with verified mentors or peers with zero health data exposed.
Anonymous connection matching
Verified mentor profiles
Zero biometric data shared
ETHICS & PRIVACY
Built Responsibly from Day One.
What We Do
All data processed on-device only
Fully opt-in, user controls everything
Evidence-based resource suggestions
Transparent about AI limitations
AI supports, it does not decide
PHIPA, PIPEDA, HIPAA-friendly
What We Don't Do
Share data with employers or third parties
Diagnose medical or mental health conditions
Replace therapists or professional care
Auto-escalate without user consent
Track without user awareness
Make choices on behalf of users
Risk Mitigation
False positives → context-aware thresholds
Over-reliance → prompt professional referrals
Stigma → private-first, no sharing defaults
Data gaps → graceful degradation
Crisis states → immediate resources surfaced
Misinterpretation → every insight explained
IMPACT
Why This Matters.
For the Engineer
Early warning Detect burnout risk before it escalates.
Full control Your data — you decide everything.
Real resources Evidence-based, not generic AI filler.
No surveillance Company has zero visibility into your health.
Lasting habits Builds self-awareness and wellbeing over time.
Future Roadmap
— Device Testing + HealthKit integration
— Opt-in anonymized team-level burnout insights
— Expanded API — GitHub, Jira, PagerDuty
For the Company
— Burnout is a major cause of on call engineer turnover
— Offer V-1 licenses as an employee perk — like gym memberships
— Shows commitment to mental health without any surveillance
— No competitor offers this: privacy-first, engineer-owned
— Reduces on-call burnout and improves engineer retention
OBC
OnCall Balance
"We are not building for cool tech —
we are building for real human wellbeing."
Privacy First
Prevention Focused
Human Centered
Jeboy · Séphora · David | CBC McGill Hackathon 2026 | Rootly Sub-Challenge
Private · Local · Powerful