newtsim is building simulation infrastructure that replaces physical prototyping and testing with computational methods that are just as reliable — and orders of magnitude faster.
newtsim was founded by a research team with backgrounds in energy storage, materials science, and computational simulation. The team combines deep domain expertise in the industries we serve with the solver engineering required to build simulation tools from scratch.
We developed the platform alongside design partners from oil & gas, pharmaceutical, and energy storage companies. Their engineers used early builds on real projects, and their feedback shaped every product we ship today. The problems our customers face are problems we have worked on together.
We develop every solver in-house, validate against published experimental data, and use the platform daily for our own research.
Physical prototyping is slow, expensive, and limits how many ideas a team can explore. Every failed prototype is weeks of lost time. Every successful one only proves a single design point.
Our goal is to make simulation accurate enough and fast enough that it replaces physical validation for the majority of the design cycle. Engineers should be able to iterate hundreds of times in simulation before a single physical test — and when that test happens, it confirms what they already know.
Longer term, we are working to deepen our understanding of the core physics that underpin multiscale simulation. As that understanding improves, the same engines that run our current products become applicable to every simulation workload — not just the three verticals we serve today.
Every newtsim product runs on the same proprietary simulation stack. We built each layer from first principles rather than integrating third-party solvers.
Our quantum chemistry engine handles electronic structure calculations for binding energies, reaction pathways, and interface chemistry. Molecular dynamics solvers handle transport properties, free energies, and conformational sampling at classical speed. Continuum solvers cover CFD and FEM for macro-scale process engineering. Data flows directly between scales — no file translation, no manual handoff.
All solvers are GPU-native, designed for massively parallel hardware from day one. This is not a CPU codebase ported to GPU. The architecture difference means parameter sweeps and design-of-experiments campaigns that would take weeks on traditional tools run overnight.
We are currently developing world models — learned representations of physical systems that combine first-principles simulation data with ML to predict system behaviour across conditions the solver has not explicitly evaluated. The goal is simulation-quality predictions at a fraction of the compute cost, enabling real-time design exploration and closed-loop automated optimisation.