// portfolio
Karim
Elbehiri
role:
Engineering intelligence at the intersection of AI and physics. Building tools where simulation meets machine learning.
// portfolio
Engineering intelligence at the intersection of AI and physics. Building tools where simulation meets machine learning.
I'm an engineer and entrepreneur with an MSc in Fluids Engineering from the University of Manchester. My thesis fused machine learning with CFD by integrating a Tensor Basis Neural Network (TBNN) directly into OpenFOAM for advanced turbulence modelling. Prior to that, I completed a BEng in Mechanical Engineering at the University of Sheffield, with my thesis about running a full-scale F1 car CFD simulations in Ansys Fluent.
At Ansys, I ran hundreds of high-fidelity simulations across heat transfer and aerodynamics, automated HPC workflows in Python, and won the annual company hackathon with a predictive ML model. Most recently, I founded MOTiON, an AI-powered sports analytics platform where I built the full iOS stack in SwiftUI, trained motion models with CreateML, and pitched to investors across the UAE.
I operate at the boundary between rigorous engineering simulation and applied machine learning, turning physically complex problems into data-driven solutions.
AI-powered sports analytics platform using Apple Watch motion sensors for real-time biomechanical performance analysis. Full iOS stack in SwiftUI with CreateML-trained motion classification models deployed on-device via CoreML.
MSc thesis: integrated a Tensor Basis Neural Network directly into OpenFOAM as a custom turbulence closure. The TBNN replaces classical RANS stress-strain assumptions with a data-driven correction trained on high-fidelity DNS data.
BEng dissertation: full-scale F1 car CFD simulations in Ansys Fluent, analysing aerodynamic drag, downforce coefficients, and wake structures at racing speeds. Mesh of 40M+ cells with k-ω SST turbulence modelling.
Automated end-to-end server room airflow simulation pipeline at Ansys using PyAnsys and HPC cluster integration. Reduced manual setup time from hours to minutes via parameterised geometry generation, meshing, solving, and post-processing.
MOTiON
Ansys
Manchester Stinger Motorsports
Hyperloop Manchester
University of Manchester
2021 - 2022 · Manchester, UK
Integrated a Tensor Basis Neural Network (TBNN) into OpenFOAM to improve turbulence closure modelling beyond classical RANS approaches combining ML with CFD.
University of Sheffield
2018 - 2021 · Sheffield, UK
Full-scale F1 car CFD simulations in Ansys Fluent analysing aerodynamic drag, downforce coefficients, and wake structures at racing speeds.
Feel free to reach out if you're looking to collaborate on an engineering problem, build an AI-driven product, or for a quick chat.