Expand your horizons with AI.




Mind Foundry are thrilled to have been included in the ‘Ethical AI startup landscape’ research, mapped by researchers at the EAIGG (Ethical AI Governance Group), who have vetted nearly 150 companies working in Ethical AI across the globe. This important research by the EAIGG is being conducted to provide transparency on the ecosystem of companies working on ethical AI. 

Mind Foundry was highlighted in both the ‘Targeted AI Solutions’ and ‘ModelOps, Monitoring & Observability’ categories. Both these subsets of Ethical AI aptly describe how Mind Foundry provides Responsible AI. 


Mind Foundry’s Targeted AI Solutions 

Mind Foundry creates Responsible AI for high-stakes applications. We build targeted AI solutions for our customers, ranging from intelligence management, insurance, central and local government, and security & defence. We emphasise the need for Responsible AI across the development lifecycle of an AI system, including:

1) Use-case specific risks: making sure our customers can succeed by fully understanding the benefits and risks of using AI for their particular business uses, and where AI should, and should not be used. 

2) Algorithmic design: favouring interpretable and explainable AI models, with data and model provenance, over black-box approaches. For example, in high-stakes applications, it is not always appropriate to use neural networks - as this can make the traceability and interpretability of your outputs opaque to users of the system as well as to unrepresented stakeholders, such as citizens.  

3) Solution design: empowering the human to make the right decision, with UX design highlighting possible limitations in the system itself. For example, by using Bayesian optimisation, we visually represent probabilistic estimates to our insurance customers, so that they can be aware of where the system is less confident in its predictions, and where additional human input might be required before making a decision. 

4) Post-deployment monitoring: ensuring our AI systems continue to work as intended through performance monitoring, including in predictive power, robustness, and resilience.


Mind Foundry’s approaches in ModelOps, Monitoring and Observability 

One of the fundamental aspects of our work is Continuous Metalearning: we are currently building and implementing the tools and techniques for the next generation of responsible AI systems. 

This research, as part of an Innovate UK Smart Grant, includes understanding how AI systems can continuously improve and adapt to surrounding environments, and meta-optimise their learning process through the combination of cutting-edge machine learning techniques and domain expert input. 

At its core, Continuous Metalearning proposes to create a complete end-to-end framework for the operation of algorithms. By prioritising these techniques, we are enabling our customers to use AI that is resilient to adversarial attacks, such as data poisoning, as well as being able to classify novel trends that the AI system had previously never seen before. 

We hope that our approaches and philosophies surrounding the development of Responsible AI continue to be spread, and we are grateful for the essential research being carried out by EAIGG to uncover this. 

Find out more about how we’re using Responsible, Explainable AI in high-stakes applications

Frankie Garcia

Written by Frankie Garcia

Frankie Garcia is a Product Manager at Mind Foundry, specialising in deploying AI in high impact situations where human-AI collaboration and ethics are imperative. Frankie's focus includes continuous metalearning: looking into how machine learning models can become future proof by continuously updating, learning how to learn, and better adapting to new environments. Frankie’s background in social sciences, philosophy and user research shaped her passion for responsible AI, and desire to create human-centred machine learning.











Upcoming Events


17 sep


25 sep


14 sep update


Human in the loop.

Sign up to get notified next time we publish.