Start as a managed characterization service that ports customer workloads, runs repeatable power and performance test suites, and delivers regression reports with threshold alerts and suspected root causes. The first product can be a standardized lab workflow: ingest workloads, define benchmarks, collect telemetry, compare runs, and produce engineering-ready triage packets. Over time, the repeatable pieces can become a hybrid toolchain plus services business for silicon validation teams.

AI accelerators, edge AI chips, and custom silicon programs are growing, while workload behavior increasingly spans models, firmware, drivers, networking, and power envelopes. Teams are hiring for this capability because generic observability tools do not solve silicon-level workload characterization.