Governed ML software for R&D teams
AUROLINK builds tools for teams working with scarce, expensive data. Our systems rank options, show uncertainty, decline unsupported inputs, and keep the evidence traceable. We started with alloy screening and use the same approach for scoped custom builds.
What we do
We build for teams that cannot use a black box. Outputs come with uncertainty, unsupported inputs are flagged, and the evidence trail stays intact.
We turn prototypes, notebooks and specialist workflows into deployed ML tools with clear limits and reviewable outputs.
Explore services →A live screening service for alloy and additive-manufacturing R&D: ranked shortlists, uncertainty and provenance.
See the product →Our approach
A useful model still has to answer basic questions: when is it unsure, when should it refuse, and where did the evidence come from? We build those checks into the system.
Every model reports measured coverage, with confidence intervals and conservative assumptions where the data is thin.
Inputs outside the evidence are refused or given wider estimates. The system is designed to avoid confident unsupported answers.
Record-level provenance, review notes and promotion decisions are kept with the output. Data curation remains human-reviewed.
Custom builds
We help teams move from experiment to usable software, then state clearly where the data supports an answer and where it does not. Work is delivered as scoped engagements, not seats.
Predictive models with uncertainty, refusal rules and clear validation results.
Internal tools that turn model outputs into ranked recommendations and reviewable evidence.
Ingestion, curation and provenance pipelines with row-level traceability and human review.
Take an existing notebook or proof-of-concept into a deployed, monitored service.
Flagship product
AUROLINK Screening helps alloy and additive-manufacturing teams decide what to make and test first. Send a candidate list or design space; receive a ranked shortlist with uncertainty, refusal flags and provenance.
We sell the screening decision, not a bare predicted number. The service is useful when ranking candidates matters more than pretending every absolute value is precise.
A list or design space: compositions and known process routes.
Each candidate is scored, checked against the evidence, and ordered by make-and-test priority.
A shortlist with uncertainty, refusal verdicts, feature attributions and provenance.
Where we're going
We start where the evidence is strongest, then expand property by property and domain by domain as real data supports it.
The alloy-screening system is available today. We are open for scoped custom builds and design-partner screening pilots.
Retained screening and private models trained on customer process data, where the data supports a stronger system.
Expansion to more alloy properties, then adjacent material families, supported by a UK team across data engineering, ML and delivery.
Clear limits
Clear limits make the output easier to use. Here is what the system is built to do, and what it is not claiming to be.
The company
AUROLINK Innovations Ltd builds ML software for materials and industrial R&D. The company is based at the Canterbury Innovation Centre, University of Kent, and operates its own alloy-screening product alongside scoped custom work.
Get in touch
Tell us what you are trying to decide. We will say what the data can support, what it cannot, and what a sensible first engagement looks like.