Internship - MLFF Distillation & GCMC Integration

About the role

CuspAI is actively seeking an engineering intern to join the computational chemistry team for a focused three-month project aimed at enhancing the efficiency of novel material discovery. The successful candidate will be responsible for developing machine learning force fields specifically designed to support high-throughput Monte Carlo simulation protocols. A core part of this role involves the technical integration of these trained models into the proprietary kUPS simulation framework used across the organization. You will work at the intersection of machine learning and molecular simulation, applying distillation methods to transfer knowledge between complex model representations. The position emphasizes close collaboration with senior scientists to ensure that the generated models meet rigorous scientific accuracy standards. This internship provides direct exposure to the end-to-end pipeline of computational materials research within an industrial setting. You will contribute to the validation and optimization of screening workflows that directly influence the speed of discovery. The work produced will have a tangible impact on the company's capacity to evaluate new molecular structures efficiently.

Key facts

• Benefits include 28 days of holiday plus public holidays, a professional development budget, and parental leave (26 weeks for primary caregivers, 12 weeks for secondary).
• CuspAI is an equal opportunity employer and provides reasonable accommodations for the interview process upon request.

What you'll do

Develop and train machine learning force fields capable of accurately representing molecular interactions relevant to industrial screening.

Implement efficient data pipelines that prepare and preprocess large-scale molecular dynamics and Monte Carlo simulation datasets for model training.

Apply advanced model distillation techniques to reduce the computational cost of high-fidelity reference simulations without sacrificing predictive fidelity.

Embed the trained machine learning models into the kUPS simulation engine to enable high-throughput virtual screening campaigns.

Conduct rigorous validation of the integrated models by comparing simulation outputs against experimental and high-level computational benchmarks.

Collaborate with data scientists and computational chemists to iterate on model architectures and loss functions based on scientific feedback.

Automate the generation of simulation reports and performance metrics to track the impact of the integrated models on workflow speed.

Participate in code reviews and documentation efforts to ensure that the implementations are maintainable and reproducible.

Support the optimization of Monte Carlo move sets to improve the sampling efficiency of the embedded force fields.

Troubleshoot numerical instabilities or performance bottlenecks that arise during the integration of machine learning models into production workflows.

Coordinate with the engineering team to align the simulation pipeline with the broader software development lifecycle.

Assist in the preparation of internal technical summaries that communicate the progress and results of the integration work.

Evaluate the trade-offs between model complexity, simulation speed, and accuracy in the context of real-world screening problems.

Contribute to the continuous improvement of the kUPS framework by proposing and testing enhancements to the model interface.

Requirements

You are currently enrolled in a graduate or undergraduate program specializing in computer science, physics, chemistry, or a closely related technical field.

You possess a strong foundational understanding of molecular simulation concepts, including Monte Carlo methods and molecular dynamics.

You have hands-on experience with at least one high-level programming language, with a demonstrated ability to write clean and maintainable code.

A working knowledge of machine learning frameworks such as PyTorch or TensorFlow is essential for this role.

You have completed coursework or projects that involve the implementation of force fields or the use of neural networks for scientific modeling.

Strong proficiency in Python is required, as the majority of the tooling and integration work will be conducted in this language.

You must be able to commit to a continuous three-month period of full-time engagement based in London.

You are eligible to work without restrictions in the United Kingdom or hold a visa sponsorship that allows for full-time employment.

A keen interest in computational chemistry, materials science, or chemical engineering is mandatory for success in this position.

Nice to have

Familiarity with common molecular simulation packages such as LAMMPS or GROMACS is advantageous.

Experience with automated force field parametrization or the development of transferable interaction models.

Knowledge of uncertainty quantification methods in machine learning applied to scientific modeling.

Background in high-performance computing or optimization for scientific workloads.

Practical notes

This is a 3 months internship based in London, UK.

You will need to be eligible to work full-time in the United Kingdom or hold a visa that permits employment.

If you require a visa to work in the UK, please note that we are unable to provide sponsorship for this role at this time.

The listed start date is flexible within the next 4 weeks, and the internship is scheduled to begin promptly.

Please ensure your application reflects your availability to start within this narrow window.

Applications will be reviewed on a rolling basis until the position is filled, so early submission is encouraged.