Numbers and methodology

Numbers and methodology

What the numbers mean—and what they do not prove.

Public counts link to their source. Company measurements are labeled internal when their benchmark, sample, or raw data cannot be published.

UPDATED JULY 20265 CASE STUDIES7 ROLES
Public or directly inspectable counts
Internal and project-reported measurements
MD78

concurrent workstreams

Internal project count
MD798%

positive feedback

Internal feedback figure; sample details are not public
Meta / Reality Labs<100 ms

reported prototype latency

Internal measurement
Intel / XeSSup to 40%

reported frame-rate improvement

Internal benchmark; workload details are not public
Intel / XeSS35%

reported inference-latency reduction

Internal benchmark
Intel / XeSS25%

reported temporal-artifact reduction

Internal evaluation
NutanixEnterprise

deployment context

Public scope is limited by confidentiality
Texas InstrumentsResearch

training and evaluation

Ongoing work
CNF Technologies15,000+

reported concurrent users

Internal production figure
CNF Technologies35%

reported p99 latency reduction

Internal production measurement
Revent$300K+

reported monthly volume

Company-reported figure
Revent99.9%

reported availability

Company monitoring figure
Method

Read every number with its source.

Public count: derived from a visible profile or the items listed on this site. Project benchmark: produced by the linked repository and not independently reproduced. Internal measurement: supplied from company work, with raw data unavailable publicly.

Overlapping roles are listed separately and are never summed into an inflated “years of experience” figure. A missing denominator or benchmark setup is treated as a limitation, not hidden behind a verification label.