Engineering / advanced
Account for AI capacity and cost
Build reproducible cost and utilization models using explicitly fictional prices, sensitivity analysis, and stated service-quality constraints.
Your route through engineering
Follow the path or begin with a question. Every lesson is open.
Your place, at your pace. Progress is optional and saved only in this browser when available.
Read / explore / reason
Inside this module
Module 01 advanced
Account for AI capacity and cost
Build reproducible cost and utilization models using explicitly fictional prices, sensitivity analysis, and stated service-quality constraints.
- 01
Resource time: preserve the unit before multiplying a rate
Translate allocation lifetimes into resource-time and hypothetical billable units without confusing wall time, kernel activity, and invoice quantities.
28 min - 02
Fixed and variable costs: specify the decision horizon
Separate shared allocation from avoidable spending and use local cost models only within an explicit capacity range.
28 min - 03
Utilization: name the useful work and the available capacity
Model committed capacity and useful resource-time while separating allocation, busy hardware, and service-feasible utilization.
28 min - 04
Inference cost: count successful logical tasks
Define a cohort and acceptance oracle before comparing costs across retries, output quality, and latency.
28 min - 05
Training cost: define useful token contributions
Separate raw processed positions from accepted objective contributions and charge replay, padding, and evaluation consistently.
28 min - 06
Storage and transfer: account for volume over time and path
Build a unit-consistent fictional model of retained capacity, operation counts, replication, and boundary-crossing data.
28 min - 07
Service frontiers: preserve quality and latency while comparing cost
Distinguish feasible configurations and Pareto dominance from unconstrained low cost in controlled AI serving experiments.
28 min - 08
Checkpoint economics: state the failure model before optimizing
Derive a first-order checkpoint interval and fictional expected resource cost while identifying the assumptions that break the approximation.
28 min - 09
Capacity sensitivity: expose the assumptions behind a crossover
Compare fictional operational capacity models with a common workload, feasible horizon, and explicit sensitivity to uncertain inputs.
28 min - 10
Cost ledgers: reconcile allocations and reserve before admitting work
Separate delayed accounting, organizational cost attribution, and real admission controls with explicit residual exposure.
28 min