Lead AI Engineer

Onto Innovation is a leader in process control, combining global scale with an expanded portfolio of leading-edge technologies that include: 3D metrology spanning the chip from nanometer-scale transistors to micron-level die-interconnects; macro defect inspection of wafers and packages; metal interconnect composition; factory analytics; and lithography for advanced semiconductor packaging. Our breadth of offerings across the entire semiconductor value chain helps our customers solve their most difficult yield, device performance, quality, and reliability issues.

Onto Innovation strives to optimize customers’ critical path of progress by making them smarter, faster and more efficient.

Onto Innovation designs and manufactures the inspection and metrology tools that the world’s leading chipmakers rely on. We believe AI will be a key enabler of faster, more reliable measurement and decision‑making in semiconductor manufacturing.

This is a hands‑on technical leadership role. You will work closely with system engineers, algorithm developers, and applications teams to understand complex metrology workflows and data flows. You will set the vision and architecture for AI in metrology, then design, build, and product ionize AI systems, including agentic ones, that speed up setup and troubleshooting and improve measurement quality. The emphasis is on agents grounded in real tools and data, with the right human oversight.

You will also set technical direction and mentor the AI and ML engineers around you.

• Set the vision and own the architecture. Define the AI strategy and reference architecture for metrology, choose what to build and what to buy, and keep the roadmap tied to the problems customers actually have.
• Build it yourself. Write, test, and ship production code — prototypes, agent workflows, evaluation tooling, and the systems that run them — and set the standard by example.
• Design agent workflows for the metrology lifecycle — guided recipe and model setup, qualification runs, tool matching, and root‑cause triage when a measurement drifts.
• Give agents real tools. Connect them to our simulators, solvers, and tool data so the agent orchestrates and the physics computes.
• Ground agents in fab knowledge: retrieval manuals, recipes, service tickets, and historical measurement outcomes, with provenance on every answer.
• Design for trust. Human‑in‑the‑loop checkpoints, confidence and uncertainty carried through every step, full audit trails, and refusal when the evidence is thin — a number without an error bar is not a measurement.
• Deploy where fabs allow it. On‑premises and air‑gapped inference, with local or hosted models chosen on latency, cost, and data‑control constraints, and a release path that survives customer change control.
• Build the evaluation harness: golden task sets, tool‑call correctness, agent‑built versus expert‑built recipes compared against the TMU budget, and regression tests on every model or prompt change.
• Keep the measurement science sound. Surrogate models, parameter inference, tool calibration, and calibrated uncertainty remain the engines agents call. Own the ML behind them where it matters, including domain shift across tools, layers, and fabs.
• Lead technically. Set standards, run design and code reviews, mentor mid‑level and junior engineers, and make the approach and build‑versus‑buy calls. Raise software quality with unit testing, test automation in CI/CD pipelines, refactoring of legacy code, and static and dynamic analysis.
• Set the practice for AI‑assisted development. Apply AI coding tools responsibly to coding, debugging, documentation, and test generation, and maintain secure, reliable workflows for evaluating…