Engineering intelligence for the physical world.
We are building AI systems that help engineers design, test and improve physical systems, starting with analog, RF and mixed-signal semiconductors.
Illustrative animation, not a measurement. Real runs record every candidate, its verdict and the inputs behind it.
Understand the requirements. Choose the next experiment. Build on experience.
Our working system connects engineering intent, predictive models, simulation and verification. It explores circuit trade-offs, guides physical layout, and brings prior design experience into new decisions. We are building toward a shared workspace where engineers and AI shape physical systems together.
Typed intent
Requirements become machine-checkable obligations.
Choose useful experiments
Prediction and uncertainty guide the next simulation.
Select against the obligations
Verdicts across process, voltage and temperature.
Build on what was learned
Compatible design history feeds search and system planning.
Semiconductor design. A working foundation.
Adopt an individual Nectrion capability or deploy the full system in your engineering environment. Validate either on a bounded design block against requirements your team sets, with replayable simulation and verification evidence.
Design-partner validations bring the system to market; verified work builds the experience record that guides later searches.
Better engineering trade-offs at the same simulation budget
On an SG13G2 LNA, Bayesian search achieved 14.6 percent higher median shared-reference Pareto hypervolume than an evolutionary baseline across five seeds at 60 simulations per arm (typical corner), taking about 1.6× the wall-clock time of the baseline.
An op-amp closed to a commercial-tool witness
End to end from a typed specification, replayable from pinned code and inputs. Internal commercial-tool witness.
An in-house electromagnetic solver
Parameter-free solver for on-chip passives, compared against Cadence EMX in a low-frequency solver comparison across 22 committed geometries.
Built from experience shipping RF silicon.
Fifteen years in RF and analog silicon
Design lead on front-end work shipped in Apple MIMO, Google Pixel and Samsung programs at Skyworks and pSemi (Murata), and most recently VP Analog Design at Sphere Semiconductor. Named inventor on two US patents.
AI/ML cofounder or founding engineer
We are looking for a complementary AI/ML cofounder or founding engineer to extend the working models, strengthen evaluation and build the shared engineer-and-AI workspace.
Bring an engineering challenge. Help build what comes next.
Start with a bounded analog, RF or mixed-signal design problem. Partner with us on the first application, or join as the founding AI/ML engineer extending the intelligence behind it.