Scientific Software Engineer

Tact; Greater London; On-site; GBP 90,000 - 130,000; Full time

Benefits offered by this job

Founding equity

Direct work with founders

Opportunity to publish work

Job summary

Tact is seeking a Scientific Software Engineer in London to build GPU-native simulation software from the ground up and train Physics AI models on data from proprietary simulations.

You will work closely with the founders to shape the platform architecture, engage with aerospace and automotive customers, and contribute to publishable results while enjoying founding equity. Strong math and coding skills are essential.

Qualifications

• PhD or industry equivalent in a computational field.
• Experience building scientific or simulation software.
• Fluency with PyTorch or JAX on physics or simulation data.

Responsibilities

• Build GPU-native simulation software from scratch.
• Train Physics AI models on data from simulations.
• Collaborate with founders to shape architecture and deliverables.

Skills

PhD in computational science

Strong coding skills

Numerical methods

Mathematics aptitude

Education

PhD in computational science

Tools

PyTorch

JAX

Job description

Scientific Software Engineer | GPU | Physics AI | PyTorch / JAX

London

Love writing fast code for hard physics problems? This is the one.

Why consider this role?

• Build GPU-native simulation software from scratch, not maintain legacy code
• Train Physics AI models on data from your own simulations
• Founding equity, so you actually own a piece of what you build
• Work directly with the founders and shape the whole architecture
• Work with their customers in the aerospace and automotive sectors
• Backing to publish your work

This is a VC-backed deep tech business in London, building a full Physics AI platform for engineering.

Their software runs physics simulations on GPUs, then trains AI models on the results so engineers get answers in seconds instead of hours.

You don't need to be a CFD specialist. They want great scientific software engineers and Physics AI people who love the maths behind it.

One week it's speeding up GPU code, the next it's training a neural operator, the next it's improving the solver.

You’ll ideally have:

• A PhD in a computational subject (or the industry equivalent)
• Built scientific or simulation software yourself, not just used it
• PyTorch or JAX experience, ideally on physics or simulation data
• Comfortable with the maths (PDEs, numerical methods, linear algebra)
• Properly strong coding skills

Bonus if you've worked with neural operators, PINNs or differentiable simulation, or contributed to open-source scientific code.

If you'd rather build the simulation engine than run it, you'll love this.