NeuralSoftX — AI for industrial electron microscopy

Custom AI and computational workflows for industrial electron microscopy.

NeuralSoftX is a specialist consultancy led by Dr Iván Lobato — physicist and developer of the MULTEM simulation platform. The company takes on focused industrial projects where generic tools or off-the-shelf models stop being reliable: rigid and non-rigid registration, tomography preprocessing, low-dose restoration, segmentation, CBED and 4D-STEM preprocessing, and simulation-driven model development.

  • 1,800+ citations across Science, Nano Letters, PNAS, PRL, Ultramicroscopy, Acta Cryst., and npj Computational Materials (h-index 21)
  • Cosslett Award — Best Invited Paper, M&M 2018
  • Developer of MULTEM — open-source TEM / STEM / CBED / EELS simulation platform, GPL-3.0, actively maintained since 2014
Illustration of industrial electron microscopy workflows

Three things NeuralSoftX does well

Workflows built for real instruments and real industrial constraints.

Robustness

Methods that survive real data

Drift, non-rigid motion, low dose, scan artefacts, and scarce labels are the norm in industrial acquisition — not edge cases. NeuralSoftX designs workflows that fail gracefully on imperfect data instead of breaking on the first artefact.

Synthetic data

Simulation as leverage

When industrial labelled data is scarce, MULTEM-generated synthetic datasets make model development possible. Physics-based simulation of TEM, STEM, CBED, and EELS is the difference between "we can't build this" and "we have a baseline in weeks."

Delivery

Deployable software, not papers

Methods are delivered as ONNX packages, on-prem inference tools, or integrations into existing microscopy pipelines — not as a notebook the client has to rebuild. The engagement ends when the workflow runs at your site.

Sectors

Industries where difficult microscopy data is the bottleneck.

Semiconductors and electronics Batteries and energy materials Advanced materials R&D labs Metals and additive manufacturing Particle, cleanliness, and contamination analysis

See industry detail

How engagements work

Start with the workflow problem — not a generic brief.

Every engagement begins with a concrete microscopy data problem: what the instrument acquires, what's failing today, and what outcome is needed. NeuralSoftX returns with a feasibility read and a scoped proposal. Engagement shapes are usually feasibility assessment, method package, deployable inference tool, or workflow integration.