Scheduling
Fit people, work, equipment and deadlines together while respecting availability, qualifications, dependencies and service levels.
// Auxil Solvers
Choose among too many options, with the choice checked independently
Turn written rules, hard requirements and competing goals into a practical recommendation that can be checked independently.
// The problem
Some decisions become difficult because each choice affects several others. A workable roster must cover the work, respect availability and qualifications, stay within budget and remain fair. A layout must fit, remain accessible, meet standards and use space well.
Spreadsheets and manual expertise can handle a few options. They struggle when there are thousands or millions of valid combinations. A solver compares those possibilities systematically and returns a recommendation together with evidence that it meets the requirements.
Those requirements often begin as policy, standards, contracts or expert guidance rather than neat system inputs. The first job is to turn that language into explicit rules without losing the judgement and exceptions in the source.
// Where it fits
Choices with constraints and trade-offs
Allocation
Distribute limited budget, capacity, inventory or staff across competing demands according to the priorities that matter.
Configuration
Choose a compatible combination of components, services or settings from more possibilities than a person can compare reliably.
Layout
Arrange physical or digital space around dimensions, access, safety, usability and capacity requirements.
// Which product?
Foundry answers. Solvers choose.
You need a consistent answer
Given this person's records and the rules that applied, what are they entitled to?
The facts already exist. The job is to apply written rules consistently, explain the result, and identify cases the rules cannot settle without human judgement.
You need to choose a good option
Given these people, jobs and restrictions, what schedule should we use?
The answer does not exist yet. Written rules define which choices are allowed and what a good result means. The solver compares the possibilities, rejects those that break a requirement, and finds a strong option according to the priorities you set.
Many business problems contain both. Written rules define what is allowed; a solver chooses the best available option within them. Auxil uses the same discipline around both: clear inputs, independent checking, traceable evidence and an explicit route for questions the software cannot decide.
// Current examples
Current examples
// Parking layout
Mosaic: parking layout
Mosaic takes a site boundary, obstacles, entrances and local design standards, then searches for a high-capacity parking layout. Every proposed layout is checked separately for dimensions, access and circulation.
Working now: site import, real-polygon geometry, circulation, a browser interface, live progress, and downloadable drawing and certificate files.
Current evidence: layouts are checked against a synthetic test suite and compared with a calculated benchmark. Commercial-product comparisons are outside the current evidence base.
// Interface layout
Interface layout
The same approach can arrange an interface around hierarchy, available space, focus order, touch targets and accessibility requirements instead of relying on a sequence of visual guesses.
Working now: the research, decision model and early prototypes needed to test the opportunity.
Current scope: engagements begin by agreeing the customer problem, baseline and independent measure of improvement.
// How Auxil approaches it
Measure improvement instead of promising magic
This is a neuro-symbolic approach in practical terms: AI helps turn natural language into a proposed set of rules; people approve what those rules mean; formal software searches within them; and a separate check verifies the recommendation. Ambiguous or undecidable cases are returned for judgement rather than forced into a confident answer.
01
Translate the written rules
AI helps identify requirements, preferences and exceptions in the source material. A person reviews that interpretation before it governs any recommendation.
02
Establish the baseline
Measure the current manual process, spreadsheet, incumbent product or expert answer so improvement has a real meaning.
03
Check every recommendation
A separate checker confirms that the proposed answer obeys the requirements. The system that creates the answer is not trusted to approve its own work.
04
State what remains unknown
The result includes what was tested, how it compared with the baseline, and what has not been established. A best-known answer is never presented as a proven best answer.
// Why not just use AI?
AI helps explore. Evidence decides.
Auxil uses AI throughout the work to understand the problem, test approaches and accelerate implementation. It is guided by the actual requirements and by experience of where apparently sensible systems fail in production.
AI can suggest a schedule, configuration or layout. It cannot prove that it considered the right requirements, found a valid answer, or improved on what you already had. That needs agreed measures, a separate check and difficult examples chosen to expose mistakes.
// Validation engagement
Test the value before funding a full product
- Bring a representative decision, real inputs and the requirements it must satisfy.
- Agree what a useful result means and what the current approach achieves.
- Run a time-boxed trial, including awkward cases and examples that should be rejected.
- Receive the recommendations, validation evidence and a clear build, revise or stop decision.
The outcome may support a product build, identify a narrower useful tool, or show that the current approach is already good enough. All three save you from investing on an untested assumption.