Search by what the work actually is

Computer Science, AI & Scientific Computing

What this division builds for the others, who builds it, and the rule that no model ships without its baseline.

Scale 10^-100 m to 10^100 m

Fluent in Python, and able to write code somebody else can read, test and extend. · Data structures and algorithms to the level of choosing correctly and justifying the choice. · Comfortable on Linux, and with Git as a working tool rather than a save button. · Has written tests for numerical code and can explain what a tolerance in a test means. + 6 more

Computer Science, AI & Scientific Computing

A scientific prediction task with a proper held-out split and a classical baseline reported beside the model.

Scale 10^-100 m to 10^100 m

Fluent in Python, and able to write code somebody else can read, test and extend. · Data structures and algorithms to the level of choosing correctly and justifying the choice. · Comfortable on Linux, and with Git as a working tool rather than a save button. · Has written tests for numerical code and can explain what a tolerance in a test means. + 6 more

Mathematics & Computational Foundations

Whole modelling problems from other divisions: formulation, analysis, and an honest statement of validity.

Scale 10^-100 m to 10^100 m

Can state and use the definition of convergence for a numerical scheme, not just cite the order. · Linear algebra to the level of conditioning, decompositions and why an ill-conditioned problem is not a bug in the solver. · Real analysis: limits, continuity, and what makes a problem well-posed. · Has implemented a numerical method from a paper and verified its convergence order empirically. + 5 more

Mathematics & Computational Foundations

A modelling question from another division formulated mathematically and checked for well-posedness.

Scale 10^-100 m to 10^100 m

Can state and use the definition of convergence for a numerical scheme, not just cite the order. · Linear algebra to the level of conditioning, decompositions and why an ill-conditioned problem is not a bug in the solver. · Real analysis: limits, continuity, and what makes a problem well-posed. · Has implemented a numerical method from a paper and verified its convergence order empirically. + 5 more

Computer Science, AI & Scientific Computing

The computation another division needs, made correct, fast enough, and reproducible.

Scale 10^-100 m to 10^100 m

Fluent in Python, and able to write code somebody else can read, test and extend. · Data structures and algorithms to the level of choosing correctly and justifying the choice. · Comfortable on Linux, and with Git as a working tool rather than a save button. · Has written tests for numerical code and can explain what a tolerance in a test means. + 6 more

Mathematics & Computational Foundations

Where the mathematics and the machine meet: conditioning, floating point, and what the computed answer is worth.

Scale 10^-100 m to 10^100 m

Can state and use the definition of convergence for a numerical scheme, not just cite the order. · Linear algebra to the level of conditioning, decompositions and why an ill-conditioned problem is not a bug in the solver. · Real analysis: limits, continuity, and what makes a problem well-posed. · Has implemented a numerical method from a paper and verified its convergence order empirically. + 5 more

Computer Science, AI & Scientific Computing

Making the department's workloads run at scale, with measured scaling curves rather than claims.

Scale 10^-100 m to 10^100 m

Fluent in Python, and able to write code somebody else can read, test and extend. · Data structures and algorithms to the level of choosing correctly and justifying the choice. · Comfortable on Linux, and with Git as a working tool rather than a save button. · Has written tests for numerical code and can explain what a tolerance in a test means. + 9 more

Computer Science, AI & Scientific Computing

Parallelising one real workload and reporting the measured speed-up and where it stopped scaling.

Scale 10^-100 m to 10^100 m

Fluent in Python, and able to write code somebody else can read, test and extend. · Data structures and algorithms to the level of choosing correctly and justifying the choice. · Comfortable on Linux, and with Git as a working tool rather than a save button. · Has written tests for numerical code and can explain what a tolerance in a test means. + 6 more

Mathematics & Computational Foundations

Building a model of a real system, stating its assumptions, and testing where it stops describing the system.

Scale 10^-100 m to 10^100 m

Can state and use the definition of convergence for a numerical scheme, not just cite the order. · Linear algebra to the level of conditioning, decompositions and why an ill-conditioned problem is not a bug in the solver. · Real analysis: limits, continuity, and what makes a problem well-posed. · Has implemented a numerical method from a paper and verified its convergence order empirically. + 5 more

Mathematics & Computational Foundations

The mathematical structure of physical theories, working across into Division 01.

Scale 10^-100 m to 10^-18 m

Can state and use the definition of convergence for a numerical scheme, not just cite the order. · Linear algebra to the level of conditioning, decompositions and why an ill-conditioned problem is not a bug in the solver. · Real analysis: limits, continuity, and what makes a problem well-posed. · Has implemented a numerical method from a paper and verified its convergence order empirically. + 5 more

Mathematics & Computational Foundations

What this division proves and builds, who does it, and the rule that a method ships with its regime of validity.

Scale 10^-100 m to 10^100 m

Can state and use the definition of convergence for a numerical scheme, not just cite the order. · Linear algebra to the level of conditioning, decompositions and why an ill-conditioned problem is not a bug in the solver. · Real analysis: limits, continuity, and what makes a problem well-posed. · Has implemented a numerical method from a paper and verified its convergence order empirically. + 5 more

Mathematics & Computational Foundations

One mathematical result reconstructed in full, including the steps its source omitted.

Scale 10^-100 m to 10^100 m

Can state and use the definition of convergence for a numerical scheme, not just cite the order. · Linear algebra to the level of conditioning, decompositions and why an ill-conditioned problem is not a bug in the solver. · Real analysis: limits, continuity, and what makes a problem well-posed. · Has implemented a numerical method from a paper and verified its convergence order empirically. + 5 more

Mathematics & Computational Foundations

Numerical methods implemented, verified against their theoretical order, and handed on as reference code.

Scale 10^-100 m to 10^100 m

Can state and use the definition of convergence for a numerical scheme, not just cite the order. · Linear algebra to the level of conditioning, decompositions and why an ill-conditioned problem is not a bug in the solver. · Real analysis: limits, continuity, and what makes a problem well-posed. · Has implemented a numerical method from a paper and verified its convergence order empirically. + 5 more

Mathematics & Computational Foundations

The optimisation methods the department depends on, and the characterisation of when each one fails.

Scale 10^-100 m to 10^100 m

Can state and use the definition of convergence for a numerical scheme, not just cite the order. · Linear algebra to the level of conditioning, decompositions and why an ill-conditioned problem is not a bug in the solver. · Real analysis: limits, continuity, and what makes a problem well-posed. · Has implemented a numerical method from a paper and verified its convergence order empirically. + 7 more

Computer Science, AI & Scientific Computing

Models that carry the physics as a constraint, and the honest account of when that helps and when it does not.

Scale 10^-100 m to 10^100 m

Fluent in Python, and able to write code somebody else can read, test and extend. · Data structures and algorithms to the level of choosing correctly and justifying the choice. · Comfortable on Linux, and with Git as a working tool rather than a save button. · Has written tests for numerical code and can explain what a tolerance in a test means. + 7 more

Computer Science, AI & Scientific Computing

How the department applies machine learning to science, and the evidence standard that applies to all of it.

Scale 10^-100 m to 10^100 m

Fluent in Python, and able to write code somebody else can read, test and extend. · Data structures and algorithms to the level of choosing correctly and justifying the choice. · Comfortable on Linux, and with Git as a working tool rather than a save button. · Has written tests for numerical code and can explain what a tolerance in a test means. + 6 more

Mathematics & Computational Foundations

The department's hardest mathematical questions and the standard every numerical claim is held to.

Scale 10^-100 m to 10^100 m

Can state and use the definition of convergence for a numerical scheme, not just cite the order. · Linear algebra to the level of conditioning, decompositions and why an ill-conditioned problem is not a bug in the solver. · Real analysis: limits, continuity, and what makes a problem well-posed. · Has implemented a numerical method from a paper and verified its convergence order empirically. + 5 more

Computer Science, AI & Scientific Computing

Machine learning on scientific data, with the evaluation designed before the model is chosen.

Scale 10^-100 m to 10^100 m

Fluent in Python, and able to write code somebody else can read, test and extend. · Data structures and algorithms to the level of choosing correctly and justifying the choice. · Comfortable on Linux, and with Git as a working tool rather than a save button. · Has written tests for numerical code and can explain what a tolerance in a test means. + 8 more

Computer Science, AI & Scientific Computing

Taking a working research script and turning it into a tested, documented package somebody else can install.

Scale 10^-100 m to 10^100 m

Fluent in Python, and able to write code somebody else can read, test and extend. · Data structures and algorithms to the level of choosing correctly and justifying the choice. · Comfortable on Linux, and with Git as a working tool rather than a save button. · Has written tests for numerical code and can explain what a tolerance in a test means. + 6 more

Computer Science, AI & Scientific Computing

One numerical method implemented, tested against an analytic case, and profiled.

Scale 10^-100 m to 10^100 m

Fluent in Python, and able to write code somebody else can read, test and extend. · Data structures and algorithms to the level of choosing correctly and justifying the choice. · Comfortable on Linux, and with Git as a working tool rather than a save button. · Has written tests for numerical code and can explain what a tolerance in a test means. + 6 more

Computer Science, AI & Scientific Computing

Research software built to survive handover: tested, documented, and maintained.

Scale 10^-100 m to 10^100 m

Fluent in Python, and able to write code somebody else can read, test and extend. · Data structures and algorithms to the level of choosing correctly and justifying the choice. · Comfortable on Linux, and with Git as a working tool rather than a save button. · Has written tests for numerical code and can explain what a tolerance in a test means. + 6 more

Mathematics & Computational Foundations

Several divisions' mathematical problems at once, and the rigour standard applied across them.

Scale 10^-100 m to 10^100 m

Can state and use the definition of convergence for a numerical scheme, not just cite the order. · Linear algebra to the level of conditioning, decompositions and why an ill-conditioned problem is not a bug in the solver. · Real analysis: limits, continuity, and what makes a problem well-posed. · Has implemented a numerical method from a paper and verified its convergence order empirically. + 5 more

Computer Science, AI & Scientific Computing

Several divisions' computational needs at once, and the engineering standard applied across them.

Scale 10^-100 m to 10^100 m

Fluent in Python, and able to write code somebody else can read, test and extend. · Data structures and algorithms to the level of choosing correctly and justifying the choice. · Comfortable on Linux, and with Git as a working tool rather than a save button. · Has written tests for numerical code and can explain what a tolerance in a test means. + 6 more

Computer Science, AI & Scientific Computing

Simulation platforms: solvers, verification, and the convergence evidence behind every run.

Scale 10^-100 m to 10^100 m

Fluent in Python, and able to write code somebody else can read, test and extend. · Data structures and algorithms to the level of choosing correctly and justifying the choice. · Comfortable on Linux, and with Git as a working tool rather than a save button. · Has written tests for numerical code and can explain what a tolerance in a test means. + 6 more