Decisions you can inspect.

Make the constraints, trade-offs and resulting placement rank visible before compute moves.

Change the objective. Read the result.

Each objective changes factor weights and candidate order. Rejected regions stay visible so the decision remains legible.

Constraint and ranking model

Deterministic policy

0 rejected

No single axis dominates. The scheduler trades cost against latency and carbon until the marginal gain on each is equal.

8/8 eligible

Paris clears every constraint at once: 6 ms, $2.19/hr and 56 g/kWh with headroom to grow.

About this decision model

The model is deterministic and uses seeded candidate data. It is not a live placement decision or provider inventory.