Leverage extrusion data for faster process decisions

The IEP is a decision-support platform for food extrusion professionals. It turns published extrusion science extracted from 300+ peer-reviewed studies into structured evidence, predictive tools, and an AI chatbot that helps you explore, reason, and act with more confidence. It is not a replacement for your expertise or your extruder; it is a way to make better-informed decisions faster.

Primarily product developers and R&D teams working with food extrusion, researchers and academics in the extrusion field, and extrusion troubleshooters looking for evidence-backed guidance. If you work with an extruder and need to make decisions about process conditions, formulation, or product quality, the platform is built for you.

A working knowledge of extrusion is helpful. The platform is designed for practitioners, not complete beginners. That said, ExtruBot can handle plain-language questions, and the prediction engine guides you through the workflow step by step, so you do not need to be a modelling expert to get useful outputs.

The platform is currently in beta. Access during the beta period is by invitation. Pricing for general availability has not yet been announced.

ExtruBot is the platform's AI chatbot, trained on a curated knowledge base of extrusion literature, symposium transcripts, and educational content. You can ask it questions about extrusion science in plain language and it will return structured, referenced answers drawn from that knowledge base.

The key difference is domain specificity and traceability. ExtruBot uses the ChatGPT API along with a curated extrusion knowledgebase, so its answers are grounded in studies that are specifically about extrusion, not synthesised from the general internet. Every claim can be traced back to a source publication, which you can open and read directly from within the platform.

Questions about how process conditions affect product attributes, ingredient-outcome relationships, troubleshooting starting points, and general extrusion science concepts. ExtruBot works best when you give it context. The more specific your question, the more useful the answer.

ExtruBot is designed specifically for extrusion science questions. Questions outside this domain will be redirected.

By default, your queries and responses are stored to allow you to review your chat history and help improve the platform. You can opt out of query logging in the ExtruBot interface. Please read the platform privacy statement on the Onboarding page before sharing any confidential or commercially sensitive information.

The prediction engine is a structured workflow that helps you identify the process conditions most likely to achieve your target product quality. You define what you want your product to be, which attributes matter, and how much, and the engine generates a set of candidate process conditions drawn from the literature, ranked by predicted quality. It then helps you design a trial plan and log your results to refine the model over subsequent iterations.

The platform is a decision-support tool, not a precision instrument. Because the underlying data comes from published literature covering many different extruders, scales, and configurations, predictions are directional rather than exact. Think of the output as an evidence-backed starting point, a smarter first trial, rather than a guaranteed result. The more you use the platform and log your own trial data, the more contextualised the predictions become for your specific setup.

At this stage the model is scale-agnostic. It does not incorporate machine-specific parameters like screw diameter or L/D ratio into the prediction. You can record your machine details in the platform for context, but these do not currently affect the model output. This is a known limitation and something that may be addressed in future versions.

For most product types, ingredient inputs are used to filter and contextualise literature data but are not yet fully integrated into the process-condition prediction. The platform will tell you clearly at each step what information is and is not being used in the model. Future development will bring tighter ingredient-process interaction modelling.

As many as you like. You can define multiple output attributes simultaneously, for example hardness, expansion index, and colour, and assign each a weighting that reflects its importance to your product. The engine optimises across all of them.

If a product attribute falls outside your defined specification range, the overall quality score for that trial is treated as failed, regardless of how well other attributes performed. The platform reflects this clearly in the results visualisation.

Yes. After running your suggested trials, you can log your actual measured results back into the platform. These are used to refine the model for your next iteration, making the predictions progressively more relevant to your specific process and product.

Yes. You can download a CSV template to record your trial results offline and upload it back into the platform, or enter results directly in the interface.

A tool that lets you search for structured claims extracted from extrusion literature, for example which studies report a relationship between screw speed and hardness, and in which direction. Results include a summary of the evidence, the number of studies supporting each direction of effect, and direct links to the source publications.

The current knowledge base includes data extracted from over 300 peer-reviewed extrusion studies, covering a range of product types, ingredients, and process conditions. The database is manually curated, with automated expansion planned for future versions.

The relationship claims were extracted from peer-reviewed extrusion studies using a curated workflow that combines structured review and data extraction. Claims about variable relationships, outcome direction, and study context are normalised into a common format so they can be searched, compared, and linked back to the original publication.

Yes. The data explorer gives you a tabular view of all studies in the knowledge base, with filtering by product type, extruder manufacturer, and other parameters. You can also plot relationships between variables, for example moisture content versus L* value, across filtered subsets of studies.

On registration you provide your name, email, role, organisation, product types you work with, and the challenges you face. Your query history and platform activity are logged by default. Please review the full privacy statement for details of how this data is stored and used.

No. Your account data, query history, and trial results are private to you. Aggregated and anonymised platform usage data may be used to improve the platform.

No. The platform operates via a silo-based learning approach. The inputs and results you add can refine the baseline model for your own project, but they are not shared with other users and are not added back into the baseline dataset to retrain the baseline model.

Exercise caution. While your data is private within the platform, you should read the privacy statement before inputting any commercially sensitive information. The platform is designed to help you reason about extrusion science, not to store proprietary formulation data.

During the beta period, access is by invitation. If you are interested in participating, contact the IEP team directly. Once registered, you will complete a short onboarding questionnaire to help tailor the platform to your product focus and challenges.

Use the feedback button within the platform to submit comments at any point. You can attach a screenshot of your current screen to give context. All feedback is reviewed by the development team.