Sr. / Staff. Engineer, Device Statistics & Model Correlation

Location: San Jose, CA (on-site)

Ayar Labs is solving the I/O bandwidth bottlenecks inherent in modern AI compute architectures. As pioneers of co-packaged optics (CPO), we develop silicon photonics solutions that deliver unprecedented bandwidth density and reach, at a fraction of the power consumption required to scale next-generation AI models.

Backed by industry giants like NVIDIA, AMD, Mediatek and Intel and manufactured in partnership with the world's leading semiconductor ecosystem, Ayar Labs' co-packaged optics solution is key to unleashing next-generation AI scale-up architectures.

Joining our Link Design and Architecture team, you will own the statistics that connect our measurements to our models. Our photonic devices are measured at wafer, package and system level, on our own test floor and by our foundry and vendor partners. You will turn that data into the statistical description of each device (distributions, variance across lots, wafers and dies, parameter correlations), prepare it for the statistical models behind our compact models and link simulations, and track how well full simulation models follow full-system measurements so discrepancies are caught early.

The team is responsible for the robustness and performance of the optical links that power our technology, from component-level modeling through architecture definition, link analysis, specification development, and yield prediction, and drives the exploration of next-generation link architectures that shape our long-term roadmap.

Essential Functions

• Work with our data infrastructure team to bring internal test data (wafer, package and system level) and foundry and vendor deliveries into a form ready for analysis, with a traceable record of where every number came from.
• Build the statistical description of each photonic device: fit distributions, separate lot-to-lot, wafer-to-wafer, die-to-die and within-die variance, estimate parameter correlations and covariance, and decide which of that structure the model needs to carry.
• Work with photonic device engineers to choose the measured parameters that best characterize each device, and with link modeling engineers to express the resulting statistics in the form the compact and link models consume (Monte Carlo and corner inputs, covariance-aware tolerancing).
• Maintain the correlation between full simulation models and full-system measurements across tapeouts and lots, track model-to-measurement residuals, and raise discrepancies with a diagnosis of whether the model or the test is at fault.
• Contribute to the device specification-setting methodology: parametric yield estimates and specification limits derived from measured distributions.
• Document methods, assumptions and data lineage so results are reproducible and reviewable by engineering teams and executive stakeholders alike.

Required Qualifications

• MS or PhD in Electrical Engineering, Physics, Optics/Photonics, Statistics, or a related field.
• 4+ years of experience analyzing measurement data for semiconductor or photonic devices.
• Strong applied statistics: distribution fitting, variance decomposition, correlation and covariance estimation, regression, hypothesis testing, design of experiments, Monte Carlo methods.
• Strong Python skills for scientific computing and data analysis (NumPy, SciPy, pandas, statsmodels or similar) and comfort working with large tabular datasets.
• Working knowledge of photonic device physics sufficient to interpret what a parameter shift means (modulators, photodetectors, grating couplers, waveguides).
• Excellent communication skills and a collaborative working style across device, modeling and test teams.

Preferred Qualifications

• Prior experience with silicon photonics or with foundry process-control data (WAT/PCM, JMP).
• Experience with statistical parameterization of compact models
• Experience validating measurement systems and test methods (gauge R&R, repeatability versus device variation).
• Experience with data pipelines and versioned datasets (Git, SQL or Parquet, dashboards).
• Familiarity with Bayesian or hierarchical models for lot, wafer and die structure.