Research Engineer

The role

Build the models that make engineering design search fast and trustworthy: prediction models trained on simulation results, the search that tunes each design within its limits, and the tests that prove a design step works before a customer relies on it.

Full simulations take hours. You build small models that screen many design options in seconds, say how certain they are, and send a shortlist to the full calculation. You measure every step on past projects whose answers are known, and say clearly when the evidence is too thin.

What you will do

• Train prediction models on simulation and calculation results, with calibrated uncertainty, so full analyses run on a shortlist.
• Build the numerical search that tunes dimensions and positions within the limits the rules set, across several objectives.
• Measure the design changes a language model proposes, and improve how it proposes them.
• Build test sets from past projects with hidden answers, and benchmark designs with known results.
• Measure how well recorded expert rules reproduce expert decisions on hidden cases.
• Report every result reproducibly, failed experiments included.

What you bring

• Strong Python and applied modelling, including surrogate models of physical simulations.
• Optimisation under constraints: multi-objective, Bayesian or gradient-free methods.
• Hands-on work with engineering simulation data such as FEA, CFD or system models.
• Disciplined evaluation: held-out cases, calibration and clear accounts of what did not work.
• A plus: a degree in mechanical, structural, aerospace or another engineering discipline.

Methods you will use

Surrogate modelling, design of experiments, Bayesian optimisation, multi-objective search, calibration and conformal prediction, and held-out case evaluation. Useful additions where the work needs them: physics-informed models, active learning and language-model evaluation.

What success looks like

A prediction model that screens design options for one package with measured accuracy against the full calculation, and a search that finds options a senior engineer accepts on a past project.

How we work

We work with clear responsibility and direct communication. We test ideas against evidence, write down important decisions and flag problems early. High standards and appreciation belong together: we expect careful work and treat the people doing it with respect.

The company

Embreier builds Kalyr, AI for industrial engineering teams. Kalyr puts senior experts’ know-how to work in every system and every team: it does routine design work in the customer’s own software, within the rules their experts set, and the customer’s engineers decide and sign. We are based in Lausanne, Switzerland.

To apply

Send your CV and two short answers through the application form. Specific answers about work you did yourself help us most.

Need an adjustment to apply? Write to careers@embreier.com and tell us how we can help.