The Federated Learning market is moving

Hey there, I'm back on the radar! The Federated Learning market is moving - let's dive into what's been happening in the last two months.

What's ahead in this newsletter:

Learning Hub:

• Navigating the tricky terrain of healthcare data privacy and European regulations

News from Apheris:

• Insights from relevant panels and summits from across the technology landscape

Federated Mind Guest Section with Nicolas Couronneau :

• What does the Flower Labs and 英伟达 (Flare) collaboration mean for the Federated Learning market?

Industry Round-up:

• Dr. Baker and Dr. AlQuraishi on the impact of AI on protein engineering
• The rising tide of PETs
• Basecamp Research 's triumph in protein structure prediction

Learning Hub

How to turn healthcare data into gold in a privacy minefield

Healthcare data is a treasure trove of insights. But have you ever tried to unlock its value? If so, you’re probably familiar with the privacy landmines in this field. In his latest interview, our CEO, Robin Roehm , explores the “healthcare data paradox” – the challenge of harnessing this wealth of information while maintaining privacy and security.

Solving the European health data puzzle in style

My son recently brought me a 1000-piece jigsaw puzzle – with no picture guide. That’s how it feels to work with health data in the midst of European regulation. Fortunately, Jan Stücke is our guide through this maze, using computational governance to piece it all together.

In his latest blog, Jan turns the array of newest regulatory pieces into a paint by numbers picture.

News from Apheris

Apheris is making waves in the tech ecosystem! Here’s a quick roundup of recent features:?

Apheris featured in Criteria Venture Tech ’s LLM Infrastructure Stack Market Map in the Federated Learning category

Apheris featured in Creandum ’s AI Governance and Security Landscape in the Federated Learning category

Apheris featured in Dealroom.co ’s Deep Tech report in the Novel AI category

Read Dealroom’s full report

AI for Good

Britta Srivas , AI Solution Engineer at Apheris, recently took part in the #AIforGood panel in London, discussing secure data access for Machine Learning (ML).

Some key insights:

• Stuart Jennings spoke about adapting media practices to maintain unbiased reporting, highlighting AI’s role in improving the way news is shared and allaying fears of AI replacing human writers.
• Amber Shafi spoke about the role of AI in drug discovery, aiming to lower the traditionally high barriers and potentially bring new treatments to market faster.
• Britta Srivas highlighted the importance of regulation and governance in AI for healthcare, noting Europe’s progress, but pointing to the need for better data capture processes to fully realize the benefits of AI.

Who's the guy pointing at this poster? Insights from the RWE Summit 2024

Caught your eye, didn't he? That’s our colleague Dan P. , deep in the heart of the IMPACT RWE Summit 2024, pointing out the future of healthcare data on a poster (aka Apheris’ solution and method).

Here are our team’s key takeaways from the conference:

• Federated Learning & Computational Governance: The summit highlighted the shift towards computational methods for securely linking diverse healthcare data.
• Large Language Models (LLMs) for healthcare: The use of LLMs to harmonize and make healthcare data interoperable was a major theme.

?

Frontiers of ethical AI at the Nordic Innovation Summit

Ellie Dobson , our beloved VP of Product, joined other technology leaders from Telenor , 爱立信 , Equinor , and 微软 , at the Nordic Innovation Summit to explore the future of ethical AI.

Their workshop highlighted:

• The Power of collaboration: Bringing together diverse expertise for ethical AI development.
• The role of decentralized data: Advocating for federated learning to improve data privacy and security.
• Governance as a pillar: Emphasizing rigorous auditability and traceability for the ethical use of AI.

Federated Minds Guest Section: About the Federated Learning market

I’m excited about this one, folks! Nicolas Couronneau , Director of AI Product at Apheris , dives into the latest movements in the Federated Learning market.

To give you an idea of why I’m so excited: Nic’s background is impressive, with a stint at Graphcore sharpening ML algorithms and experience in developing algorithms for autonomous vehicles before his time with us. He truly understands the ins and outs of Federated Learning and the broader ML market. Having recently hit up major events like Paris.AI and the Flower Summit, Nic brings back fresh, insightful perspectives on AI and the evolution of Federated Learning.

“Federated Learning is beginning to move from being a field of scientific exploration to delivering real business value. This shift is marked by the introduction of FLOps – the operationalization of Federated Learning within a product environment. The concept of FLOps focuses on the development of Federated Learning platforms that are secure, scalable, stable, and performant, with the versatility to handle different computing requirements, users, and workloads. In addition, these platforms must be auditable and upgradeable to meet the rigorous demands of real-world applications.

The last few weeks have seen a collaborative effort to improve Federated Learning’s enterprise capabilities. It’s been my personal highlight to see Flower gaining traction during the Flower AI Summit. The summit saw the exciting announcement of a closer collaboration between Flower and NVFlare. Partnerships such as these are vital for driving the adoption of FL in enterprise settings.

However, exploring Federated Learning in an enterprise context brings to light an important realization: Federated Learning is often indispensable for guaranteeing data privacy, but is not a stand-alone solution. For wider adoption of Federated Learning, it’s critical to consider the end-to-end aspects of governance, privacy, and security concerns to ensure a responsible approach.

The protection of the intellectual property (IP) of AI models has consistently been at the forefront of discussions within the Federated Learning community. The rationale for this focus is clear: the development and training of AI models not only requires substantial resources, but also represents a significant capital investment. With the commoditization of AI infrastructure, ensuring the confidentiality of trained models throughout the entire ML pipeline becomes critical. Federated Learning coupled with confidential computing – as demonstrated by NVIDIA’s Blackwell platform – represents a promising path to secure Federated Learning environments. This applies not only to the training of models but also to cross-site model evaluation: the idea of assessing the performance of a model on confidential data, possibly belonging to a different organization, whilst protecting the model IP from the data owner.

The way forward for Federated Learning is one of collaboration and an unwavering commitment to privacy and security. Apheris, with its Federated Learning core and strong governance, privacy, and security framework, has clearly solidified its position as a leader in this space. It has become increasingly clear that in regulated industries, where the protection of sensitive data is a concern, the governed, private, and secure use of Federated Learning is the way forward.” (Nicolas Couronneau)

Industry Round-up

What's the future of protein design? Dr. David Baker and Mohammed AlQuraishi on the impact of AI on protein engineering

Following groundbreaking developments such as AlphaFold2 and RoseTTAFold, a conversation in Nature Biotechnology brings insights from Dr. David Baker and Mohammed AlQuraishi on accelerating the impact of AI on protein engineering. Their dialogue illuminates the journey from structural prediction to tangible advancements in drug discovery.

Explore their insights

Proprietary data as a secret ingredient for success

Basecamp Research ’s BaseFold, has just outpaced AlphaFold2. How? Through the power of their unique dataset, #BaseGraph.

PETs left the cute stage

Privacy enhancing technologies (PETs) are on the brink of their "hockey-stick moment," as Dr. Rand Hindi points out. With the rapid expansion of AI and big data, the demand for privacy “tools” such as homomorphic encryption, federated learning, and zero-knowledge proofs is skyrocketing.

That's it for this month!

Cheers, Marie Roehm