I basically never hear the fan of my Macbook Pro, but if I do, I know whose fault it is

My favorite shirt has a great quote on it: “Without Data, you’re just a person with an opinion.”

If I look around, I see a lot of people with opinions, especially here.

So, it’s actually refreshing that someone rolled up their sleeves to change that and ask leaders and practitioners what’s hype and what’s real when it comes to Data & AI, and what their challenges and priorities are.

We collected, analyzed, and distilled these data in the Data & AI monitor.

I think it turned out great.

Thank you for your attention to this matter!

https://pages.xebia.com/artificial-intelligence/data-ai-monitor-report-2025-2026

No

Being a parents brings you joy, responsibilities, and boring stuff.

Thanks to LLMs, the list of boring stuff just got a bit shorter for me—and I wrote about it on our blog!

hashtag#llm hashtag#kids hashtag#automatetheboringstuff

Rob Dielemans 😉

https://xebia.com/blog/automate-the-boring-parts-of-parenthood-with-llms/

I always thought that in the summer, everyone’s is vacationing and it doesn’t make sense to run any courses.

But, last year, the team proved me wrong: they run a successful Data Summer School.

Since one data point is not a trend, they want to prove me wrong again: they’ve organized an entire week dedicated to learning Data & AI.

And they’ve been thorough, with a program with:

✅ Data Science Foundation ✅ Data Science Next Level ✅ Generative AI

All tracks are designed to be practical, with up to 80% immersed in exercises designed to empower you to learn today, apply tomorrow.

Curious to find out more? Lucy Sheppard shares a glimpse of what the Summer School looks like: https://www.youtube.com/watch?v=idU9CpBrajA

If you want to join us in Amsterdam on August 11-15, or if you know someone who would benefit, share or get in touch!

💥 And, last but not least, a special summer pricing (up to 40% off) is available.

🔗 https://xebia.ai/dss-2025-gl

hashtag#DataSummerSchool hashtag#DataScience hashtag#MachineLearning hashtag#GenerativeAI hashtag#Python hashtag#Learning hashtag#Amsterdam

Shared Links

• https://www.youtube.com/watch?v=idU9CpBrajA

External Links

• https://www.youtube.com/watch?v=idU9CpBrajA

Yesterday, I attended the AWS GenAI for Healthcare Summit in Lisbon.

Here’s my impressions for the day, and thanks Krishna Singh and Jens Dommel for inviting me!

Where is the Medical Sector using (Gen)AI

A non-exhaustive list:

1. Clinical Decision Support
2. AI Buddy in Diagnostics
3. Enhancing Imaging For example, using AI to adjust X-ray imaging when there’s missing data, low resolution, cropping, or low contrast.
4. Data Storage Management Some institutions have kept all their digital X-ray images. Since storage is so costly they started leveraging AI to discard 70% of their data and reconstruct when needed from the remaining 30% without information loss.
5. Reducing Invasive Examinations Research suggests up to 5% of all cancers may be linked to exposure from CT scans. To avoid it, less precise scans can be used, making diagnosis harder. TheraPanacea trained a variational autoencoder on 250,000 CT scans to generate high-quality images from lower-quality scans. The demo results were impressive!

I found it fascinating that 3–5 are “human-out-of-the-loop” applications—offering scalable, software-like gains!

The most moving example, however, was a highly “human-in-the-loop,” non-scalable solution. In Africa, there’s a shortage of radiologists and widespread tuberculosis. To help, a new portable X-ray box with built-in AI models is being used. It’s transported village-to-village by motorbike. Everyone in the village can be scanned, and if the AI flags someone’s scan, they’re referred to a hospital for specialized care. The motorcyclist doesn’t need medical training—the X-ray box and AI do the work!

hashtag#Challenges

Customized/fine-tuned foundational models for the medical subdomain is a must. That is, not only should models be specialized in health care: they should be specialized in the discipline within health care where they will be deployed. Why?

1. Medical accuracy (goes without saying
2. Clinical relevance
3. Direct access and quoting medical sources and evidence
4. Healthcare context understanding (steering the model towards prioritizing clinical info—which is something the models should be able to do already, but I suspect that there’s a lot of improvised medical advice on the internet that a non-specialized model won’t be steered enough by prompting alone)
5. A didactic approach in the replies geared towards what health professionals are used. I cannot validate this, but I’ve heard that general models tends to be terser in their replies by default. I don’t know by which measure prompting can address it!

All these items contribute to trust, one of the big challenges in deploying AI in the medical sector. And with trust, adoption follows, and, from adoption, impact.

To close: everybody was screaming agents but there was no agentic demos

Maybe next year?

Next up at the AWS GenAI for Health Summit:

Nickolas Papanikolaou on How to increase translation from Research to the Clinics using hashtag#genai

Excited to be at the AWS hashtag#GenAI for Health Summit in Lisbon!

The day started with Andrea Rockall talking about Transforming Healthcare Through Generative AI

Radiology is a BIG area where AI can make a huge difference

The limits of LLM remain beautiful. This happened to me today and… Imagine if it happened in production to you!

What a spectacular view from Google Cloud’s Space Oddity room in London.

Had a great fun talking about human-in-the-loop in the era of Agentic AI and the role of continuous learning.

Lots of great interactions with the audience and we had so much fun and insightful conversations!