Product validation & evidence
The evidence behind
the experience.
Chart provenance, demonstration limitations and recorded verification scope. Numetiva is a product of K21SiliconSystemsPrivateLimited.
01 / Image provenance
What you are seeing.
Application scene figures
The six application-data views use hash-verified chart scenes from the final 1501 evidence catalogue. Coordinates, fitted response values, reference lines, category labels, wafer identities and recorded display sampling are retained. The website draws them as resolution-independent SVGs with larger labels and a stacked phone layout. They are vector presentations of saved application data, rather than screenshots of native controls. No model was recalculated and no die was invented. The website viewer zooms and pans a static figure; it does not execute a fitted model.
Python showcase figures
The nine Python showcase figures were generated in the local Python environment from a user-provided engineering demonstration dataset containing 1,000 records and 65 columns. They are reproducible scientific illustrations, not captures of native application controls. The original SVG figures are served without raster re-encoding. The hero is a vector detail crop of the complete Python wafer map, available on the semiconductor page. Dataset factory origin, customer identity and production performance have not been independently verified.
Reading a chart
Units, legends, axes, model labels and displayed sample information remain part of each chart. A visual reference curve does not establish normality; a profiler shows a fitted model, not causality. Capability, control limits and modelled PPM require the sampling plan, assumptions and engineering context recorded with the original analysis. Example process results are not product performance benchmarks or customer acceptance claims.
Source identifiers and preservation
Application build SHA-256: 0325e3fc6b1bdcbac0b6ebc6b6b24d8a503a2fcbffa96819a949c6477f6f6c36. Python demonstration input SHA-256: fa2333e92c955cdb1bf04dd6313d9836b11a8a82d89a8a0acc3fa8828bac705f. Source scene and original figure hashes and rendering operations are included in the owner’s website source package. Website 1503 does not modify the qualified desktop application.
Product-tour motion
Workflow nodes, moving pointers, code reveals and the AI tour illustrate selected workflows. They are not recordings of a live factory connection or current AI inference, and do not represent measured response speed. Pause motion at any time; reduced-motion preferences are respected.
02 / Recorded verification
Evidence.
With a clear scope.
The following is a summary of recorded engineering checks for application candidate 1501, compile-r25. The delivered application includes an offline validation report. These checks are finite evidence, not a promise that every possible input or environment is defect-free.
| Area | Recorded coverage |
|---|---|
| Desktop & chart routes | 26 desktop test groups; 177 Modern-theme scenarios covering 31 Graph and 138 Analyse menu routes, including 12 fitted-profiler families. |
| Rapid chart interaction | 2,976 browser debugging-protocol wheel events on visible Cartesian panels, with zoom/Fit recovery checks. This is a defined automated stress sequence. |
| Numerical checks | 26 numerical groups. A separate 39-test NIST foundation suite includes certified fits for Norris, NoInt1 and NoInt2, plus parser and boundary checks across 11 official files. This does not certify every method against NIST. |
| Invalid data | 276 empty or unavailable-numeric input cases across 138 routes: 237 expected rejections and 39 reviewed valid categorical, planning or missing-aware outputs. Source data preservation was checked. |
| Measurement systems | Five MSA modes tested with a 1,000-row fixture; the 250-row timeout regression was addressed. Independent covariance-reference checks were recorded. |
| Charts & reports | 131 native fixtures and five regression-recovery cases. Fourteen Office/PDF artifacts and 22 rendered pages reviewed using LibreOffice and PDF rendering; this is not a native Microsoft Office lifecycle test. |
Qualification boundaries and standards
Code signing, an independent clean-PC installer/upgrade/repair/uninstall cycle, physical 100%/150%/200% display checks and formal release approval remain separate qualification activities. A genuine authorised tester-STDF file paired with an independent tabular export was unavailable; synthetic parser checks do not replace that fixture. Licensed clause-by-clause standards review and an independent licensed JMP/Minitab comparison have not been completed.
Public references inform the applicability register; the website does not claim ISO 9001, ISO 11462, AIAG/VDA, JEDEC or AEC certification, approval or component qualification. Such standards depend on the organisation, process, applicable edition and product context, not just software output.
Website checks are separate
This website package records its own page, responsive layout, asset, navigation, gallery, keyboard, dialog, contact-draft and reduced-motion checks. Loopback timings on a development PC do not predict public hosting or customer hardware speed. The website test report is included in the source package.
03 / Local operation
Your environment.
Your controls.
Local and air-gapped use
Statistical calculations, configured offline AI and bundled Python workflows can run on the Windows PC. Optional database connections, external services, user scripts and dependency installation have their own permissions and network requirements. Air-gapped deployment requires the appropriate packages, models, configuration and network policy; no independent air-gap or security certification is implied.
AI & Python
AI assists with drafting and reviewing scripts. Free-form code correctness, repeatability and latency are not guaranteed. Check the selected columns, dependencies, transformations, method assumptions, execution errors and returned output before engineering use. The example script assumes the named numeric columns, sufficient distinct finite values and a suitable fit. The website AI button opens a product tour, not an AI service.
Connected workflows
Refresh and monitoring depend on the configured source, credentials, driver, polling settings and source availability. Website motion is illustrative; it is not a live production feed. Recipe and batch results still require input and output review.