Reports in your format, not ours.
Every extraction produces publication-grade HTML reports with live charts — generated by the AI, grounded in the actual simulation data, and blunt about what isn’t good enough yet. And the format is yours to define.
Drop in your old report
Have a sign-off report your QA already accepts? Drop it on AutoPDK. It reads the structure — sections, plot types, tables, verdict conventions — and generates the same report for the new extraction, improved where the data allows and populated with the new results.
Or just describe it
“Add gm/Id vs Vg per corner, a Vth-vs-L scaling page, and a red/amber/green summary at the top.” That sentence is the spec. Custom metrics, custom diagrams, custom layout — integrated into the report, interactively or through the API from your own systems.
Physics-grounded, honest
Every number in the report is traced to a real simulation of a real model card generation — never inferred by a black box or an ML surrogate. A bad starting card is called a bad starting card; a fit that regressed a metric is shown, not smoothed over.
Initial Assessment Report
Survey of the measurement set, card-vs-data coverage, initial fit quality with pass/fail per category, largest deviations, and a recommended extraction workflow.
Final Extraction Report
Before/after overlays for the corner devices, accuracy improvement per metric, optimized parameter summary, stage convergence, and an overall verdict with recommendation — in your template, or in ours.
Bring your report template.
Send us the sign-off report your QA already accepts — the first extraction comes back in that format.