Engineering / research
Measure energy across the AI lifecycle
Define measurement boundaries, power dynamics, cooling overhead, and emissions factors before evaluating scheduling or hardware efficiency claims.
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 research
Measure energy across the AI lifecycle
Define measurement boundaries, power dynamics, cooling overhead, and emissions factors before evaluating scheduling or hardware efficiency claims.
- 01
Name the electrical boundary before reporting energy
Convert power traces into energy while preventing overlapping device, node, and facility accounts.
27 min - 02
Distinguish gross energy from activity above a baseline
Use scoped idle and switching models without treating utilization as a universal wattage predictor.
27 min - 03
Check what a power sensor observes and how often it changes
Identify stale averages, missed bursts, cumulative-counter resets, and incomplete device coverage.
27 min - 04
Optimize completed work rather than the smallest wattage
Measure runtime and system energy under a supported power policy while enforcing workload and service constraints.
27 min - 05
Use PUE within its facility and accounting boundaries
Allocate shared overhead transparently without treating an average ratio as marginal cooling savings.
27 min - 06
Match electricity factors to the question being answered
Distinguish average accounting, marginal response, time weighting, and gas-scope conventions.
27 min - 07
Define useful output before dividing energy by tokens
Keep accepted output, processing volume, quality, and completed throughput aligned with the measured interval.
27 min - 08
Compare feasible schedules with all declared movement costs
Include deadlines, migration, storage, factor uncertainty, and demand changes in conditional shifting claims.
27 min - 09
Allocate manufacturing impact without confusing it with avoidance
Expose lifecycle scope, utilization conventions, uncertain lifetimes, and replacement break-even assumptions.
27 min - 10
Publish an energy result another experiment can challenge
Match workload and instrumentation, preserve raw evidence, and separate repeatability from systematic uncertainty.
27 min