Auditoria Algorítmica
Our service to evaluate solutions & release the potential of Trustworthy AI
It is about making your organization more competitive
We identify AI systems pitfalls that risk your operations, increasing algorithm accountability.
We deliver all necessary recommendations to move towards a competitive AI solution aligned with the European AI-Act.
What we review
Internal Validation & Model Selection
We scrutinize the AI development and implemented validation schema, reviewing aspects such as model selection and feature engineering, usually linked to overfitting issues.
External Validation & Real World Test
We perform an experimental test of the AI system within real-world conditions. We obtain an accurate assessment of the actual AI performance, and its impact on your operations when embedded in.
Explainability and Transparency
We assess why & how the AI system makes its predictions; the visibility you have on it; reduce the dangerous notion of black box AI; and facilitate the treatment of errors.
Reproducibility and Reliability
Two pillars of scientific innovation. We enable you to understand what AI systems do and how stable they are. Reliable AI works properly within a known range of inputs and situations.
Fairness & Bias
We look for biases that jeopardize AI operation. Pitfalls are usually a consequence of data drifting and changes between training and operational data inputs.
Other Data Considerations
Purpose limitation, proportionality, data minimization, privacy, ownership, regulation: the data pipeline is key for trustworthy AI. GDPR compliance.
Com ho fem
- Scope
The review process starts with an alignment to establish what is the operational context and actual use of the AI solution, and which aspects we focus the review on. - Review
The review proceeds in two distinct -but connected- steps. First, we perform a thorough review of the available documentation on the AI solution development and operation, identifying the aspects that lack enough proof, potentially hazardous to the AI operation. - Test
Second, we design and perform a real world test of the AI solution. We design the experiment to maximize its significance and the information you can extract from it. The test aims to stress the AI solution to detect pitfalls, and leverage these to improve the system. - Improve
We distill the findings to provide you with actionable insights. We provide an evaluation and scoring of the AI solution, together with recommendations to address the review findings, aiming to ease your path towards an improved AI solution.
This service is for you if…
You want a safe deployment plan that maximizes AI impact
You need to respond to regulatory bodies and e.g. get your medical certification
You have invested time & money in a solution that is not meeting expectations
You want to adhere to a responsible use of AI systems
Our Partners
