← Capabilities
LLM inference cost and latency optimisation
not yet earnedI can account for token spend per run, cut time-to-first-token against a stated budget, and say which optimisations were not worth their complexity.
What this claim costs
- 5 assessed items at tier 2 or above0 of 5 — 5 more to do, from 41 available
- A public artifact built for this claim — cost-and-latency-studyNot built yet
- A documented negative result — where this stops workingNothing recorded. A capability with no known failure boundary has not been explored.
- Defended aloud, against questions you did not write2 available, none done
- Recalled cold, a week later, without notes12 scheduled, none done
The proof
cost-and-latency-study
A streaming product profiled end to end, with a caching and routing change measured against a before baseline rather than estimated.
What it has to contain
- Token spend accounted per run, reconciled against the provider's billing
- Time-to-first-token measured against a stated budget
- One optimisation predicted quantitatively before it was applied, then measured
- An optimisation rejected as not worth its complexity, with the number
Where it stops working. Not established yet. Anyone can ship a demo; the boundary is the part that is worth reading, so this claim is not earned until it is written down.
Skills underneath
- Token economics
- Prompt and result caching
- Batching and throughput
- Model serving
- Latency budgets and streaming
- Quantisation and compression
- GPU execution model
- Distributed training
- Reliability patterns
41 assessed items across the specified courses evidence these, 0 done.