D ONE – Data Driven Value Creation

Workflow Wednesday Do You Still Sketch Your Data Pipelines? ✏️➡️🚀 Code-Driven Workflows Scale, Diagrams Don’t A photo of a crowded event—how many people are in it? A phone call transcript—who spoke when? Not long ago, these were “unstructured” data points. Today, AI extracts faces from images and timestamps from conversations, turning them into structured datasets. The line between structured and unstructured data has blurred, yet many teams still draw their data transformations in visual tools—then jump to a “free-hand SQL” section when complexity strikes. Instead of drawing workflows, modern platforms treat transformations as code, often generated dynamically to handle ever-changing data. Turning photos into lists of people or transcripts into structured records isn’t a one-off task—it’s ongoing processing for all incoming data, so insights flow continuously into analytics and AI. Ten years ago, we saw that dependency management was key to scaling. With no off-the-shelf solution, we built our own code-based pipelines. Today, tools like dbt provide these features out of the box: transformations as code, a clear dependency graph, and support for both SQL and Python. Code-driven pipelines also integrate seamlessly with CI/CD, data testing, and lineage tracking, ensuring every change is tested, documented, and deployed. Minibatches keep data fresh without the overhead of full-streaming, while automated tests guard against silent failures. It all adds up to a unified, AI-ready data environment that’s engineered rather than hand-sketched. So, are you drawing your data workflows—or engineering them? Drop a comment and share what excites—or worries—you most about this shift. 👇 #WorkflowWednesday #DataEngineering #AI #CI/CD #DataTesting #DataLineage #MakingSenseOfData Art by @basilonmypizza: https://lnkd.in/eF8FkWzN https://basilhefti.ch/

George Verouchis and Thanos Gkinakos presented an Azure OpenAI demo at Microsoft, where they uploaded volumes of invoices from pdfs and images into structured data in Azure SQL within seconds, thus supporting the procurement department Thank you for inviting us! And cool talk by Dimitris Agagiotis on how Fabric can help organizations streamline data management, enhance analytics, and drive better decision-making! #greece #azure #openai #azuresql #doneai

We ran the Lycabettus Run for the third year in a row, celebrating its 10th anniversary. Great to be back! 🏃💨💨 Congratulations to Ada Papanakli, Zoi Doxanidi, Alexandros Polychronopoulos, Marianna Arvaniti, Thanos Gkinakos and Aliki Ntouzgou for completing 5 kilometres and to Simone Totino, Konstantinos Zeakis, Stamatis Stefanakos, Konstantinos Pittas, Michael Kalligas, Panagiotis Langaris for completing 11 kilometres! 🎉 Very cool! #running #teamsport #team #doneai

We were asked about more information on our white paper about Information Permeability after the talk of Andrei Dmitrenko at AMLD last week. Here you go, link to the white paper, authored by Andrei and Moritz Haag: https://lnkd.in/dauDPNQv The slides are linked above. We really believe that information permeability translates well into healthcare and it’s a future we are working towards with our clients. If you are interested in a chat about it, please reach out to Andrei! #informationpermeability #healthcare #informationflow #doneai

Kantonsspital Baden AG talked about Isolde at AMLD - Applied Machine Learning Days - very interesting! KSB has the mission to enhance infection prevention. We built them an AI-powered chatbot that won an award. It’s called Isolde. Isolde is an AI assistant that makes medical knowledge on hospital hygiene available by chat and draws its knowledge (with sources cited) from the collected documents written by the hygiene specialists. This enables employees who are not specialized in hygiene to get the right information directly. It was important to ensure quick, precise and reproducible answers that are aligned with internal guidelines and which references original documents, enhancing users’ confidence. Dr. Benedikt Wiggli, Lead Physician for Infectiology and Infection Prevention at KSB, won the 1st Prize in Innovation in Infection Prevention with Isolde at the Swiss Society for Hospital Hygiene’s annual meeting. Thank you so much for having a talk at our “AI in Healthcare” track together with Marco Petrella! And thank you for this strategic partnership! You can read more about Isolde on our website here: https://lnkd.in/dCNBKy4N #healthcare #chatbot #hygiene #infectionprevention #partnership #doneai

Trend Tuesday: Could AI Catch a Fly? 🦟💨 A fly zips through the air with just 150,000 neurons, consuming only microwatts of energy. It stabilizes mid-flight, dodges obstacles, and reacts faster than you can swat. Meanwhile, large AI models require vast computational resources, scaling energy consumption with every generated token. Are we building intelligence the wrong way? I remember the first time I discovered SQLite—a tiny, self-contained database that runs in under 1MB. It wasn’t just small; it was powerful. It made me question why so much software is bloated. More is not always better. The same principle applies to AI. DeepSeek’s Mixture of Experts is already proving that activating only the necessary parameters saves energy without sacrificing performance—much like how (biological) neurons fire in bursts rather than all at once. 🔹 Tiny but Mighty: Efficiency Wins Smarter design has outperformed brute force before: ✅ Micro-Max – A chess engine with 2000 ELO in just 1,337 bytes. ✅ smollm – A pip-installable tiny LLM that gets the job done. ✅ LZ4 Compression – Super-fast, super-lightweight. Also musicians know: Minimalism isn’t weakness—it’s mastery. 🔹 AI’s Future: Smarter, Not Bigger Scaling up has worked, but we’re already hitting energy constraints. The future belongs to models that achieve more with less. We don’t need ever-larger models; we need more efficient architectures: ⚡ Sparse, modular AI that activates only what’s needed. 🧠 Neural architectures inspired by biological brains. 🛠️ Specialized hardware optimized for efficiency (like the experimental neuromorphic chips). Of course, AI and a fly's brain serve different purposes, but nature proves intelligence doesn’t require massive energy consumption. Instead of chasing scale, let’s chase elegance. The AI race will soon shift from "who’s bigger?" to "who’s smarter?" 💡 Where else do you see efficiency making the difference? - The wonderful art as always by @basilonmypizza: https://lnkd.in/eF8FkWzN https://basilhefti.ch/ - The fly brain has been mapped, btw, see Nature volume 634, pages 139–152 (2024) https://lnkd.in/egmpVcxt - I met R. D. Hipp of sqlite at the Tcl conference 2008 in Vancouver, https://lnkd.in/ebDr2YXD

A huge thank you to everyone who joined our track, Transforming Business Operations with AI at AMLD - Applied Machine Learning Days 2025! We truly appreciate the insightful discussions, thought-provoking questions, and engaging conversations that made this session such a stunning success. AI is reshaping the way businesses operate and generate value from data. It was inspiring to explore its impact together with our amazing speakers Thomas Peetz from Swiss Re, Stefano Santinelli from localsearch (Swisscom Directories AG), Ronald Zankl: from Implenia, Matthias Braendle from die Mobiliar and Elias Rüfenacht from Ypsomed AG and our host Joël Ineichen from D ONE – Data Driven Value Creation. Looking forward to continuing the dialogue and driving AI innovation forward! #AMLD2025 #AI #ML #doneai

Theory Thursday The Future of AI: Here, but Unevenly Distributed 🚀 “The future is already here—it’s just not evenly distributed.” – William Gibson Everett Rogers’ Diffusion of Innovations Theory (1962) reminds us that breakthroughs spread in stages—early adopters get the benefits first while many lag behind. Today, three vivid examples highlight this uneven progress: DeepSeek’s Spike-Based Processing (Mixture of Experts) 🤖 DeepSeek takes inspiration from biology by activating only the “experts” needed for a task—similar to how neurons fire selectively. This approach, echoing S.B. Laughlin’s 1981 findings in Vision Research, maximizes efficiency and cuts costs. It’s a reminder that sometimes, focused intelligence beats sheer power. Battery Innovation for a Solar-Powered Future 🔋 Utility-scale battery farms like those at Moss Landing (California) and Hornsdale (Australia) store solar energy to stabilize grids during peak demand. While these projects deliver cleaner air and enhanced energy security, they’re still confined to a few pioneering regions—leaving many areas dependent on fossil fuels. ETH Zurich’s Breakthrough in Breast Cancer Metastases 🎗️ Researchers at ETH Zurich have discovered a way to eliminate circulating breast cancer cells before they form metastases by “tricking” them into self-destruction. Although this promising therapy faces regulatory and R&D challenges, it could revolutionize cancer treatment—showing how life-saving innovations take long before reaching the masses. What Does This Mean for AI & Innovation? 💡 Innovation is accelerating - and its benefits often arrive in pockets. While cutting-edge labs like DeepSeek push the boundaries, many organizations still grapple with basic AI adoption due to mindset, limited resources, outdated infrastructure, or regulatory hurdles. Key Takeaways: 🔗 Collaboration is key: Cross-disciplinary efforts spark breakthroughs ⚡ Efficiency wins: Moving from “more compute” to “smarter compute” is essential 🚀 Act now: With top institutions like ETH Zurich, Switzerland leads—if we bridge the gap from lab to market. Not everyone sees the future at the same time—but together, we can close the gap. How do you see these innovations reshaping our world? Drop a comment or send me a message—I’d love to hear your thoughts! #AI #Innovation #DeepTech #Energy #Healthcare #FutureTech The D ONE x-mas card talks about how big ideas take time: the https://lnkd.in/eJbXVbNE Art by @basilonmypizza: https://lnkd.in/eF8FkWzN

Jacqueline Stählin will be speaking at the “AI in action” event hosted by Elevate Tech and SIX. The event is about how AI is transforming industries, with leading women in the field sharing their expertise and insights. You can listen to engaging talks on #scaling machine learning and AI-powered data access in an event that will showcase the women driving change in the world of AI and data science. Date: March 27, 2025 Time: 17:30-21:00 Location: SIX Convention Point Will you join us there? Women in Machine Learning & Data Science Zürich #ai #womenintech #womeninai #ml #doneai

We are at AMLD - Applied Machine Learning Days - come visit us! The D ONE team is open for discussions about AI/ML and forging new connections. We’d appreciate you dropping by! And keep an eye out for our tracks: Tomorrow (Wednesday): AI in Healthcare, 14:00-17:30 🏥 Thursday: Transforming Business Operations with AI, 09:00-12:30 🏢 Would be cool to see you there :) #AMLD2025 #doneai