State the governing equation
𝒩[u](x) = f(x), x ∈ Ω
ℬ[u](x) = g(x), x ∈ ∂Ω
A differential operator on the domain, with conditions on its boundary.
Scientific Research and Engineering eXcellence
Scientific machine learning, grounded in mathematics.
ScireX is an open-source framework for scientific computing and machine learning, built for researchers and engineers who need a robust, efficient and mathematically rigorous platform.
Open-source release coming soonBuilt by ZenteiQ
Physics-informed network, training liveTraining
−u″(x) = f(x), x ∈ (0, 1)
u(0) = u(1) = 0
exact: u = sin πx
Step0
Loss–
Relative L² error–
24 tanh neurons, Adam, 64 collocation points. Written for this page to illustrate the method; not ScireX source.
Physics enters through the loss function
A physics-informed neural network is trained to satisfy the governing equation itself, not only to fit observed data. The panel above runs exactly this procedure.
Represent the solution with a network
u(x) ≈ uθ(x)
θ = {Wℓ, bℓ}
Automatic differentiation gives the derivatives of uθ exactly at any point.
Minimise the physics-informed loss
ℒ(θ) = 1Nr∑i|𝒩[uθ](xi) − f(xi)|²
+ λ 1Nb∑j|ℬ[uθ](xj) − g(xj)|²
Residuals at collocation points xi and boundary points xj, weighted by λ, reduced by gradient-based optimisation.
Because the equation is part of the objective, the network is constrained by the physics at every collocation point, including where no measurements exist.
Core capabilities
Scientific computing, machine learning and research tooling in one framework, so experiments move from equation to result without changing environments.
Scientific computing
- Physics-informed neural networks
- Differential equation solvers
- Scientific data analysis
Machine learning
- Classical machine learning algorithms
- Deep learning models
- Custom loss functions
- Advanced optimisers
Research tools
- Experiment management
- Result visualisation
- Benchmark datasets
- Performance metrics
Designed for rigorous research
The principles that shape how ScireX is built and maintained.
Research-first design
Built specifically for scientific computing and research workflows.
Mathematical foundations
A strong emphasis on mathematical correctness and theoretical grounding.
Hardware-aware optimisation
Detects the libraries available on your system and configures itself accordingly.
Open source
An evolving community of users and developers, with clear documentation requirements that keep the code robust.
Scientific visualisation
Post-processing tools that make complex results straightforward to inspect.
Application ready
A flexible, data-agnostic framework that can be applied across projects.
Built and maintained by ZenteiQ
ScireX is designed, built and maintained entirely by ZenteiQ, connecting theoretical research with real-world implementation.
Designed, built and maintained by
ZenteiQ develops physics-native foundation models from scratch, combining scientific research, traceable data pipelines, model training and enterprise deployment capabilities.
Official release coming soon
ScireX has not been officially released yet. The source code and documentation will follow with the release.
DocumentationGuides and reference material
Coming soonSource codePublic repository
Coming soon