Machine Learning — PhD Intern (Modeling and Data Analysis)

Overview Keysight is at the forefront of technology innovation, delivering breakthroughs and trusted insights in electronic design, simulation, prototyping, test, manufacturing, and optimization. Our ~16,800 employees create world-class solutions in communications, 5G, automotive, energy, quantum, aerospace, defense, and semiconductor markets for customers in over 100 countries. Learn more about what we do. Our award-winning culture embraces a bold vision of where technology can take us and a passion for tackling challenging problems with industry-first solutions. We believe that when people feel a sense of belonging, they can be more creative, innovative, and thrive at all points in their careers. About the Program Keysight's AI team builds adaptive engineering intelligence systems where measurement, simulation, and machine learning converge - models that learn from complex physical data and adapt to new design and testing conditions. This internship focuses on machine learning surrogate models for engineering simulation and measurement. High-fidelity simulations and physical measurements can dominate the design and test loop for RF, high-speed, and other engineering problems, and their cost limits how much of a design space an engineer can explore. You will develop models that predict engineering responses at a fraction of the cost, investigate how such models can be adapted efficiently as the design problem changes, and integrate the result into an engineering ML runtime environment. Role Overview You will design, train, and evaluate neural surrogate models that predict engineering responses from simulation or measurement data far faster than repeated high-fidelity simulation or physical characterization. This spans the full arc of the problem: generating or assembling the data, choosing and evaluating model architectures and output representations, establishing where accuracy holds and where it degrades, and quantifying the result against a realistic simulation or measurement budget. In the second phase you will study how a trained model can be specialized efficiently when the design problem shifts, using LoRA and related parameter-efficient adaptation methods. You will work alongside AI researchers, simulation and measurement experts, RF designers, and runtime engineers. Learning Outcomes Apply modern machine learning to realistic engineering simulation and measurement data. Learn how surrogate model accuracy depends on data quality, sampling strategy, representation choices, and simulation budget. Develop practical experience with parameter-efficient adaptation methods such as LoRA under data-scarce engineering conditions. Build fluency with uncertainty estimation, physical consistency checks, and engineering-relevant error analysis. Take a research model from Python experimentation toward robust C++ and LibTorch integration. Collaborate across machine learning, RF design, simulation, measurement, and runtime engineering disciplines. Responsibilities Generate parametric engineering datasets using simulation and/or measurement sources, with attention to sampling strategy and to numerical noise in the underlying data. Design, train, and evaluate neural surrogate models for engineering response prediction. Investigate how the choice of output representation affects accuracy, particularly across operating regions or frequency ranges where the response varies sharply. Establish rigorous baselines and quantify accuracy as a function of simulation or measurement budget. Evaluate LoRA and related parameter-efficient methods for specializing trained models to new problem settings, against full and partial fine-tuning. Develop calibrated uncertainty estimates so that predictions can be trusted or deferred appropriately. Verify physical consistency and engineering constraints - including passivity, causality, and reciprocity where applicable - and report engineering-relevant metrics including worst-case behavior. Analyze the simulation corpus and model behavior: parameter sensitivity, response dimensionality, data quality and solver noise, and where prediction error concentrates across frequency and design space. Prototype model training and inference integration in C++ and LibTorch within an engineering ML runtime environment. Qualifications Required Qualifications Currently pursuing or recently completed a PhD in Machine Learning, Computer Science, Applied Mathematics, Electrical or Electronic Engineering, or a related technical field. Hands-on experience across multiple neural architectures - transformers, neural operators, graph networks, CNNs - and sound judgment about which suits a given problem. Strong Python and PyTorch, with experience training and debugging neural networks well beyond tutorial scale. Strong C++ and hands-on LibTorch experience to implement and integrate models into a production runtime. Solid grasp of transfer learning, fine-tuning, and parameter-efficient methods such as LoRA. Strong applied statistics and exploratory data analysis, including sensitivity analysis and dimensionality reduction, and the judgment to diagnose a model from its errors rather than its aggregate score. Experimental discipline: rigorous baselines, careful ablations, experiment tracking, reproducible results. Willingness to learn the underlying physics rather than treat it as arbitrary numerical data. Mentorship will be provided. Preferred Qualifications Background in RF, microwave, high-speed digital, or test and measurement engineering; familiarity with S-parameters, filters, or interconnect. Experience with EM simulation tools and their scripting interfaces, such as Keysight ADS, RFPro, Momentum, or comparable tools, or engineering measurement and test automation environments. Experience with surrogate modeling, operator learning, physics-informed machine learning, or scientific machine learning generally. Experience with uncertainty quantifi