Six capabilities that make the loop work
CRUD is infrastructure. The intelligence layer is the product. These are the six capabilities that turn a database of courses and jobs into something a planner can act on.
Skill Demand Graph
One graph relating industry, sector, role, skill and proficiency — and connecting each skill onward to the courses, modules, trainers and students associated with it.
- Industry → Sector → Job Role → Skill → Proficiency on the demand side
- Skill → Course → Curriculum Module → Institution on the supply side
- Skill → Student → Gap on the candidate side
- Modelled relationally for the MVP, behind services that could migrate to a graph store without changing callers
Skill Drift Detection
Knowing that SQL is in demand today is useful. Knowing that it is rising while spreadsheet-only analysis is falling is what lets a curriculum change land before the gap opens.
- Immutable frequency snapshots per measurement period
- Consecutive periods diffed to classify emerging, growing, stable and declining
- Emerging requires both large relative growth and meaningful absolute presence
- Obsolescence requires sustained decline across multiple periods, never a single data point
Curriculum Twin
A machine-readable representation of what a course actually teaches, so alignment can be computed rather than estimated in a committee.
- Modules with sequence, theory hours, practical hours, topics and assessment method
- Skills taught per module, with the proficiency a graduate reaches and contact hours spent
- Equipment and trainer requirements attached to the course
- Versioned, so curriculum change can be tracked over time
Skill Gap Engine
Gap analysis that accounts for importance and market reality, not just presence or absence.
- Proficiency distance weighted by how essential the skill is to the role
- Live demand score folded into severity, so high-demand gaps outrank low-demand ones
- Works at both district scale and individual candidate scale from the same model
- Readiness percentage weighted by importance rather than a flat count
Employer Validation Loop
The mechanism that keeps the intelligence honest and makes it improve with use.
- Employers confirm or reject detected demand in their own district
- Declared hiring intent is weighted above inferred posting signals
- Validated signals raise confidence on every dependent recommendation
- Rejected signals are excluded from scoring
Explainable Recommendations
A recommendation without evidence cannot survive a curriculum committee, a budget review, or a sceptical student.
- Every recommendation states its supporting numbers in plain language
- Confidence and impact are reported, not hidden
- Keyed so regeneration updates rather than duplicates
- A planner's decision to dismiss or implement survives the next pipeline run