Architecting the ShiftOS Scheduling Engine
Architecting the ShiftOS Scheduling Engine
Architecting the ShiftOS Scheduling Engine
ShiftOS is a workforce operating system for healthcare teams.
Scheduling is its core engine - the system everything else depends on.
My work focused on designing this scheduling engine: the workflows, interactions, intelligence layer, and visual system that allow hospitals and pharmacy networks to staff safely, efficiently, and predictably.
ShiftOS is a workforce operating system for healthcare teams.
Scheduling is its core engine - the system everything else depends on.
My work focused on designing this scheduling engine: the workflows, interactions, intelligence layer, and visual system that allow hospitals and pharmacy networks to staff safely, efficiently, and predictably.
Raised $3.5 Million
Raised $3.5 Million
demoing ShiftOS Scheduler by Topology Ventures
demoing ShiftOS Scheduler by Topology Ventures
demoing ShiftOS Scheduler by Topology Ventures
Landed Kroger & Ascension
Landed Kroger & Ascension
as partners for this platform
as partners for this platform
as partners for this platform
What I Built
What I Built
ShiftOS Scheduling is the module that sits at the center of healthcare operations.
My work focused on building the scheduling engine, the surface where every staffing decision begins.
The goal wasn’t to redesign a calendar.
It was to design a system that blends:
manual control for decisions that require human judgment
AI-driven reasoning for decisions that require speed, outreach, and constraint logic
consistent actions across the entire interface
predictable workflows across departments and facilities
This is the module future ShiftOS features will rely on.
Why This Needed to Exist
Why This Needed to Exist
Healthcare staffing breaks down because reality changes faster than tools can respond. Decades of legacy systems, designed around billing and compliance rather than clinical workflows, have grown into $66 Million Administrative Burden They don't understand:
Healthcare staffing breaks down because reality changes faster than tools can respond. Decades of legacy systems, designed around billing and compliance rather than clinical workflows, have grown into $66 Million Administrative Burden They don't understand:
last-minute call-offs
overtime and fatigue rules
licensing and certification requirements
PRN workflows
multi-facility assignments
EHR-specific training
cascading coverage gaps
last-minute call-offs
overtime and fatigue rules
licensing and certification requirements
PRN workflows
multi-facility assignments
EHR-specific training
cascading coverage gaps
It's all the same
Legacy tools only display the schedule.
They don’t understand the rules, conflicts, or ripple effects underneath.
Legacy tools only display the schedule.
They don’t understand the rules, conflicts, or ripple effects underneath.
Legacy tools only display the schedule.
They don’t understand the rules, conflicts, or ripple effects underneath.
Designing the System Model
Designing the System Model
Designing the System Model
The Goal was to design a behavioral system and what the scheduler is likely trying to accomplish.
The Goal was to design a behavioral system and what the scheduler is likely trying to accomplish.
The Principle
A system that studies patterns, senses intent, and updates its actions based on what you’re likely trying to do next.
A system that studies patterns, senses intent, and updates its actions based on what you’re likely trying to do next.
To achieve this, I designed the engine around three architectural layers:
To achieve this, I designed the engine around three architectural layers:
Holly - Systemwide Agent
Holly - Systemwide Agent
Interaction Layer
Interaction Layer
Logic Layer
Logic Layer
Intelligence Layer
Intelligence Layer
01 Interaction Layer
01 Interaction Layer
01 Interaction Layer
This is the surface the scheduler touches.
It captures micro-intent: what shift they clicked, what they hovered, what timeframe they’re working in, and what decision they’re setting up.
This is the surface the scheduler touches.
It captures micro-intent: what shift they clicked, what they hovered, what timeframe they’re working in, and what decision they’re setting up.
When the system senses a time-sensitive coverage need
When the system senses a time-sensitive coverage need
If the assignment requires rapid resolution, the Interaction Layer tightens into a focused, high-confidence recommendation.
Holly surfaces the best-fit staff member and asks whether she should initiate outreach.
If the assignment requires rapid resolution, the Interaction Layer tightens into a focused, high-confidence recommendation.
Holly surfaces the best-fit staff member and asks whether she should initiate outreach.
When there is ample flexibility to plan
When there is ample flexibility to plan
When the coverage window allows more space, the Interaction Layer expands into a broader, ranked list of staff members.
This supports thoughtful evaluation instead of urgency.
When the coverage window allows more space, the Interaction Layer expands into a broader, ranked list of staff members.
This supports thoughtful evaluation instead of urgency.
02 Logic Layer
02 Logic Layer
02 Logic Layer
Once an action is triggered on the surface, the system needs a way to determine what is actually allowed.
The logic layer handles this reasoning. It interprets the user’s intent against operational constraints such as licensing, fatigue limits, eligibility, facility access, training requirements, and department needs.
Once an action is triggered on the surface, the system needs a way to determine what is actually allowed.
The logic layer handles this reasoning. It interprets the user’s intent against operational constraints such as licensing, fatigue limits, eligibility, facility access, training requirements, and department needs.
Choosing the Rule Type
Choosing the Rule Type
The underlying behavior the user is trying to formalize - whether it’s about coverage, availability, limits, or compliance.
The underlying behavior the user is trying to formalize - whether it’s about coverage, availability, limits, or compliance.
Set the Conditions
Set the Conditions
The real-world boundaries that define when that behavior matters as schedules shift.
The real-world boundaries that define when that behavior matters as schedules shift.
Context-Aware Naming Assist
Context-Aware Naming Assist
Context-Aware Naming Assist
As the rule’s logic comes together, the system reads the selected conditions and suggests a concise name that reflects them.
It’s a small UI nudge that shows the system understands the structure you’ve created - without interrupting the flow.
As the rule’s logic comes together, the system reads the selected conditions and suggests a concise name that reflects them.
It’s a small UI nudge that shows the system understands the structure you’ve created - without interrupting the flow.
03 Intelligence Layer
03 Intelligence Layer
The intelligence layer is what determines how the AI sidebar responds.
When a shift is selected, the system looks at the conditions behind it - timing, who’s assigned, constraints, recent changes - and shapes the sidebar to reflect that state directly.
There’s no preset template.
The sidebar changes because the underlying context changes. It shows what’s relevant for this moment and leaves out what isn’t.
This keeps the sidebar from behaving like a static panel.
The intelligence layer is what determines how the AI sidebar responds.
When a shift is selected, the system looks at the conditions behind it - timing, who’s assigned, constraints, recent changes - and shapes the sidebar to reflect that state directly.
There’s no preset template.
The sidebar changes because the underlying context changes. It shows what’s relevant for this moment and leaves out what isn’t.
This keeps the sidebar from behaving like a static panel.
Big Wins
Big Wins
As the Lead Product Designer for the scheduling engine, my decisions shaped how the product evolved and how the company moved forward.
As the Lead Product Designer for the scheduling engine, my decisions shaped how the product evolved and how the company moved forward.
Raised $3.5 Million demoing ShiftOS Scheduler by Topology Ventures
Raised $3.5 Million demoing ShiftOS Scheduler by Topology Ventures
Helped organizations like Kroger Pharmacy, Ascension, and multi-site hospital groups evaluate and partner with ShiftOS
Helped organizations like Kroger Pharmacy, Ascension, and multi-site hospital groups evaluate and partner with ShiftOS
