News and guides

News and guides from the lab.

The content hub collects company news and a growing set of guides on surrogate modeling, coating development, and chemical process scale-up, written to explain the methods plainly before any product pitch.

Company news

Featured · Event

CADFEM Conference Rapperswil, Switzerland 2026

What happens when AI moves from a general trend into real engineering workflows? More than 400 engineers, simulation experts, and R&D professionals explored that question at the CADFEM Conference Rapperswil 2026, held at OST Campus Rapperswil-Jona on 18 June. At the invitation of CADFEM Switzerland, PI Probaligence joined as an exhibitor and speaker to show how Agentic AI supports practical engineering work.
PI Probaligence at the CADFEM AI Conference Singapore 2026Event Discover STOCHOS Flow: free online seminarEvent First prize at the Robert Bosch Foundation Health AwardCompany Introducing STOCHOS Flow: AI-assisted workflow automation for engineersRelease PI Probaligence achieves official Ansys Technology Partner statusCompany Article in Konstruktionspraxis: Automated Verification Processes at GEMÜ with STOCHOSPress PI Probaligence Visits CADFEM SEA in SingaporeEvent PI Probaligence Joins Lacktagung 2025 in Aachen, GermanyEvent PI Probaligence at the CADFEM Konferenz 2025 in RapperswilEvent AI Beyond the Buzz: How Engineers Apply AI in Real-World ProjectsPress New Office in the CADFEM BuildingCompany High Fidelity Digital Twins Using AICase study Case study: multifidelity analysis of a traversed induction hardening deviceCase study PI at the CADFEM Conference France 2025Event AI in coatings development: insights from ECS 2025Event Our interview in the expert survey: How AI is revolutionizing productionPress PI Probaligence expands its presence in the USAEvent Making repositories and AI systems usable in everyday nursing care (KIP)Company STOCHOS 6.0.0 releasedRelease CADFEM Conference 2024 AustriaEvent WOST Conference 2024Event Days of Digital Technologies, 07.-08. October in BerlinEvent PI Probaligence: From AI startup to a global playerCompany CADFEM Blueprint about AI: Answers to open questionsPress The resource-efficient advantages of the probabilistic machine learning software STOCHOS for dedicated simulation applicationsPress PI to attend the European Automotive & Plastic Coating EventEvent PI Probaligence participated in the 2nd Workference for Smart PaintEvent BioTechX webinar in partnership with AnsysEvent 'Universal machine learning based on probabilistic intelligence' at the CADFEM Conference 2024 in Rapperswil, SwitzerlandEvent Read our article 'AI for Engineering' in the engineering journal Der KonstrukteurPress Very strong interest in our ML solution at the CADFEM Conference 2024Event DLR to present the partner project with SGL Carbon and PI Probaligence at the DPG Annual Conference and Spring MeetingEvent PI Probaligence to present at the High Performance Engineering Solutions 2024 in Brasov, RomaniaEvent Project meeting with the project partners at the Fraunhofer Institute in BremenEvent PI Probaligence becomes part of the CADFEM GroupCompany STOCHOS 5.0.2 released: almost 50% speed boost with backend software upgradeRelease We at PI believe that highly effective, specialized AI solutions for specific problems are the way to goCompany Benchmark of PI-BO by ZF Friedrichshafen at the NAFEMS seminar on AI and machine learning in CAE-based simulationCase study Green Tech kick-off event for project Na, LogischEvent Results after four adaptations with partner Adler LackeCase study

Insights

12 short reads

Shorter than a guide. Each answers one question about a STOCHOS or STOCHOS Flow feature, drawn from our Technical Insights series, and links down to the page that owns the detail.

STOCHOS Flow episode card: getting your data ready without the headache
Workflow

Merge messy engineering data without scripting

The Smart Data Loader turns a plain request into an editable pandas script that merges spreadsheets, logs, and exports into one table.

 STOCHOS Flow episode card: clean data in seconds
Workflow

Clean lab data for machine learning

The Smart Preprocessor turns plain-language cleaning instructions into a reviewable sklearn and pandas pipeline you can rerun.

 STOCHOS Flow episode card: AI models that know what they don't know
Uncertainty

Know when your model is guessing

A probabilistic model returns a mean plus uncertainty bands, so you see where predictions rest on data and where it is extrapolating.

 STOCHOS Flow episode card: what is actually driving your results
Sensitivity

Which inputs drive your results

A model can predict well and still explain nothing. Sobol indices rank the inputs that drive your results; SHAP explains each prediction.

 STOCHOS Flow episode card: smarter experiments, fewer trials
Optimization

Run fewer experiments, learn more

One noisy result can send your next run to a fake peak. DIM-GP discounts it, and Bayesian optimization picks the point that teaches the most.

 STOCHOS Flow episode card: expensive experiments, only when they matter
Multi-fidelity

Combine cheap and expensive data

Multi-fidelity models learn the correlation between data sources, so dense cheap data gives the shape and scarce accurate runs anchor it.

 STOCHOS Flow episode card: when built-in is not enough
Workflow

Custom steps without writing the code

An AI assistant drafts your custom Python step from a plain description. You review the code, then export it as a reusable node.

 STOCHOS Flow episode card: describe it, see it
Workflow

Charts from a sentence

Smart Plot turns a plain-language description into editable matplotlib code, previews the chart, and exports PNG, SVG, or PDF.

 STOCHOS Flow episode card: from results to report
Workflow

From analysis to report in minutes

The AI Report Generator drafts a structured report from your workflow, captions figures, and exports to PowerPoint, Word, PDF, or LaTeX.

 STOCHOS Flow episode card: teaching AI what your team already knows
Workflow

Teach the AI your own documentation

Knowledge bases index your docs, code, and PDFs on your machine, so any AI node writes against your real APIs instead of guessing.

 STOCHOS Flow episode card: share a model, not a manual
Workflow

Share a model, not a manual

The Export Web App node turns a workflow into a standalone Streamlit app, so colleagues run your model in a browser with no Python.

 STOCHOS Flow episode card: from visual workflow to production code
Workflow

From visual workflow to production code

A STOCHOS Flow workflow exports as a standalone Python package, so the same pipeline that ran on the canvas runs in scripts or CI.

Guides

10 guides
Featured guide

Bayesian optimization for engineers

Testing every combination is not an option. See how Bayesian optimization chooses each next experiment from the last result, so you reach a better design in tens of runs, not hundreds.

Read the guide →
Guide

When can you trust an AI prediction?

How an uncertainty-aware model tells you where to trust the result and when to go back to the solver.
Guide

Build an ML workflow without an ML team

Train and deploy a predictive model with the engineer who already owns the problem.
Guide

Which of your parameters actually matter

Rank what really drives the response before the next campaign is planned.

Engineering

Predict a full field, not a single number

Peak stress does not tell you where it is. Field-capable surrogate models predict the whole distribution across the geometry in seconds, so you keep the spatial answer.
Paint

Match a color in two batches, not seven

Every correction loop costs a batch. A model trained on your own recipes proposes a closer first match, so you converge in a couple of batches instead of a week of tweaks.
Engineering

Generate design variants that respect your constraints

Manual variant building limits how many designs you ever evaluate. A model learns geometry from your existing designs and generates new ones that keep the features you cannot move.
Chemistry

Keep up with a high-throughput lab

Automation moved the bottleneck from the bench to planning. A model chooses the next plate from the last one, so throughput turns into information instead of volume.
Chemistry

Cut lab to production scale-up risk

Lab results rarely survive plant scale. Multi-fidelity models combine cheap small-scale data with a few expensive runs to predict plant behavior before the batch is made.
Workflow

Run engineering AI without the cloud

Your simulation data is your IP. A local, probabilistic AI stack trains and predicts on your own infrastructure, including fully air-gapped operation with an offline license.

Put a method to work

Request a demo and we will show STOCHOS and STOCHOS Flow on a problem close to yours.

Request a Demo Explore the science
Partners, customers, and research collaborators
AnsysCADFEMSimuTech GroupMEScoTSNENAFEMS MemberBoschZFGEMUDLRAdler LackeMankiewiczDuluxPlixxentFraunhoferHochschule NiederrheinFUELL Lab AutomationHumotionUniversitaet HamburgRobert Bosch Stiftung