Scheduler Systems
A Constraint-Solving Engine for Workforce Scheduling
Not a better UI for building schedules. A better algorithm.
Scheduling looks simple on the surface. Assign people to shifts. Fill the gaps. Hit publish.
But anyone who’s actually built a schedule for a 20-person team knows the truth: it’s a constraint satisfaction problem. Every employee has preferences. Every shift has capacity limits. Every assignment affects every other assignment. One bad decision cascades into a week of complaints, no-shows, and turnover.
26% of hourly workers have quit a job because of their schedule. The cost isn’t just operational — it’s $3,000 to $5,000 per hourly worker in turnover alone.
The tools that exist today digitize the spreadsheet. They move the grid to an app. But they don’t solve the underlying problem: how do you produce a schedule that is provably fair, operationally valid, and built on real data — not guesswork?
That’s what Scheduler was built to solve. Not with a better UI. With a better algorithm.
Origin
An Algorithm, Not an App
Scheduler didn’t start as a scheduling app. It started as an algorithmic challenge.
Shay Panuilov, the founder, is a software engineer who spent years watching managers at retail stores, restaurants, and hospitals struggle with the same problem every week: build a schedule that doesn’t make someone quit. The existing tools all approached it the same way — let the manager drag and drop shifts in a calendar. Better UI, same broken process.
Shay approached it differently. He asked: what if you modeled scheduling as a constraint satisfaction problem and let an algorithm solve it?
The result is a multi-phase constraint solver that takes ranked employee preferences, station capacities, shift rules, and fairness metrics as inputs — and outputs an optimized schedule that maximizes preference satisfaction while respecting every hard constraint.
This isn’t a recommendation engine. It’s not a suggestion. It’s a solver.
The Engine
How the Algorithm Works
Under the hood, Scheduler constructs a 7×3 priority matrix — seven days, three shifts (morning, afternoon, night) — with station-level capacity for each cell. Then it runs a multi-phase optimization:
Phase 1: Priority Processing
Every employee submits ranked preferences — not binary availability, but ordered choices. The system parses these into a structured priority map with occurrence counts per employee.
Phase 2: Matrix Construction
Builds a multi-dimensional assignment matrix: days x shifts x stations. Each cell has a capacity constraint and a set of candidate employees ranked by preference strength.
Phase 3: Fairness-Optimized Assignment
The core solver. Iterates using findPerson() (lowest fairness deficit) and findCell() (highest net benefit). Every assignment validated against hard constraints before commit.
Phase 4: Iterative Resolution
Cycles through priority layers, resolving conflicts where multiple employees compete for the same shift. Each iteration improves the global fairness score until convergence.
Hard Constraints (validated before every assignment):
- Same-day constraint — no employee works multiple shifts in one day
- Consecutive-shift constraint — no back-to-back shifts within a 24-hour window
- Station capacity — respects staffing requirements per station
- Weekly load limit — maximum 6 shifts per employee per week
The output: a complete weekly schedule for a multi-station team — generated in seconds, with a fairness guarantee that no manual process can match.
Why This Matters
UI Problem vs. Computation Problem
Every scheduling tool on the market — When I Work, Deputy, Homebase, 7shifts — treats scheduling as a UI problem. Better drag-and-drop. Better mobile app. Better notifications.
Scheduler treats it as a computation problem:
| Traditional Tools | Scheduler |
|---|---|
| Manager assigns shifts manually | Algorithm assigns shifts optimally |
| Availability is binary (yes/no) | Preferences are ranked (1st, 2nd, 3rd) |
| Fairness is subjective | Fairness is measurable (fulfillment score) |
| Conflicts found after publishing | Constraints validated before assignment |
| Scales with manager effort | Scales with compute |
This is the same class of problem as airline crew scheduling, operating room allocation, and university course assignment — domains where algorithmic approaches replaced manual methods decades ago. Workforce shift scheduling is the last holdout.
AI-Powered
Gemini-Assisted Configuration
The algorithm handles schedule generation. But before it runs, the schedule needs configuration — shifts, stations, hours, roles.
Scheduler uses Gemini-powered AI to handle setup. A manager describes their business in plain language: “I run a coffee shop with 12 employees, open 7am to 11pm, two barista stations and one register.” The AI extracts scheduling parameters automatically — shift hours, station count, staffing requirements — and pre-fills the configuration.
Setup that used to take 15 minutes now takes 30 seconds. The manager reviews and adjusts. Then the constraint solver takes over.
Infrastructure
Production-Grade Stack
50,000+ downloads. Two years of production data. Real teams in retail, hospitality, healthcare, logistics, and construction.
Native iOS, Android, Web
Real-time sync, Cloud Functions
Clean business logic separation
Data protected everywhere
Scoped visibility per role
Intelligent setup & config
Scheduler Systems: Intelligent Resource Allocation
The constraint-solving engine at the heart of Scheduler solves a specific instance of a general problem: intelligent resource allocation under constraints. Allocating people to shifts. Allocating tasks to teams. Allocating permissions to agents.
Scheduler
Allocates people to shifts — optimizing for fairness, preferences, and operational constraints. The workforce scheduling engine.
GAL (Governance Agentic Layer)
Allocates permissions to AI agents — enforcing security policies, compliance rules, and configuration standards across engineering organizations. gal.run
Different domains. Same computational DNA. The vision is an ecosystem where intelligent scheduling, project orchestration, and AI agent governance share the same foundational approach: algorithms that allocate resources fairly, under constraints, with full transparency.
Workforce scheduling was the first problem we solved. It won’t be the last.
What’s Next
- AI-generated schedules — the constraint solver proposes complete schedules from historical patterns and team preference data. Manager reviews and approves.
- Predictive analytics — preference satisfaction scores, coverage trend analysis, and turnover risk indicators.
- Bidirectional calendar sync — full two-way sync with Google Calendar, Apple Calendar, and Outlook.
- API access — expose the scheduling engine for integration with payroll, HR, and workforce management platforms.
Start Scheduling in Under 5 Minutes
Free for small teams. Forever. No credit card required.
Small teams — forever
Priority scheduling, PDF exports, unlimited team
scheduler-systems.com ·
gal.run ·
Contact
Scheduler Systems builds intelligent resource allocation systems. Scheduler optimizes workforce scheduling. GAL governs AI agent deployment. Same engine, different domains.