Andreas Horn

𝗔 𝘆𝗲𝗮𝗿 𝗮𝗴𝗼 IBM 𝗯𝘂𝗶𝗹𝘁 𝗗𝗼𝗰𝗹𝗶𝗻𝗴. 𝗔𝗻𝗱 𝗶𝘁’𝘀 𝘀𝗹𝗼𝘄𝗹𝘆 𝗯𝗲𝗰𝗼𝗺𝗶𝗻𝗴 𝗼𝗻𝗲 𝗼𝗳 𝘁𝗵𝗲 𝗺𝗼𝘀𝘁 𝘀𝘂𝗰𝗰𝗲𝘀𝘀𝗳𝘂𝗹 𝗼𝗽𝗲𝗻-𝘀𝗼𝘂𝗿𝗰𝗲 𝗽𝗿𝗼𝗷𝗲𝗰𝘁𝘀 𝘁𝗼 𝗰𝗼𝗺𝗲 𝗼𝘂𝘁 𝗼𝗳 𝗜𝗕𝗠. It's now part of the LF AI & Data Foundation and already embedded in thousands of AI pipelines worldwide (it stands at nearly at 40k GitHub stars and is pretty much constantly trending on Git or Hugging Face). 𝗦𝗼 𝘄𝗵𝗮𝘁 𝗶𝘀 𝗜𝗕𝗠 𝗗𝗼𝗰𝗹𝗶𝗻𝗴? Language models — no matter how advanced — often struggle to accurately interpret, extract, or contextualize information from documents. IBM Docling helps with exactly that. It's an open-source toolkit that converts complex documents like PDFs, Word files, and slide decks into structured, machine-readable formats such as JSON and Markdown, making them ready for use with generative AI models and retrieval-augmented generation workflows. 𝗜𝘁 𝗵𝗲𝗹𝗽𝘀 𝘆𝗼𝘂 𝘁𝗼: → Parses PDFs, Word, PowerPoint, Excel, HTML, images, audio, and more → Extracts tables, formulas, code, layouts, and metadata with precision → Runs fully local for sensitive or air-gapped environments → Exports into clean formats: Markdown, HTML, JSON, DocTags → Plug-and-play with LangChain, LlamaIndex, Haystack, Crew AI, and MCP servers Instead of brittle OCR scripts or cloud-locked APIs, Docling gives enterprises a unified, extensible way to convert unstructured content into LLM-ready data. It’s fast, private, and also designed for agentic AI workflows. And the setup is done very easily in a few lines. I’ve already recommended this to several companies - and the feedback has been nothing but excellent. Try it now: https://lnkd.in/e4E_aRtn 📬 𝗪𝗮𝗻𝘁 𝗺𝗼𝗿𝗲 𝗹𝗶𝗸𝗲 𝘁𝗵𝗶𝘀? 𝗝𝗼𝗶𝗻 𝟮𝟬𝗸+ 𝗼𝘁𝗵𝗲𝗿𝘀 𝗹𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗳𝗿𝗼𝗺 𝗺𝘆 𝗻𝗲𝘄𝘀𝗹𝗲𝘁𝘁𝗲𝗿. 𝗜 𝘀𝗵𝗮𝗿𝗲 𝘁𝗵𝗲 𝗯𝗲𝘀𝘁 𝘄𝗲𝗲𝗸𝗹𝘆 𝗱𝗿𝗼𝗽𝘀 𝗮𝗯𝗼𝘂𝘁 𝗔𝗜 𝗮𝗴𝗲𝗻𝘁𝘀, 𝗲𝗺𝗲𝗿𝗴𝗶𝗻𝗴 𝘄𝗼𝗿𝗸𝗳𝗹𝗼𝘄𝘀, 𝗮𝗻𝗱 𝗵𝗼𝘄 𝘁𝗼 𝘀𝘁𝗮𝘆 𝗮𝗵𝗲𝗮𝗱 → https://lnkd.in/dbf74Y9E

531

74 Kommentare

As part of our strategic partnership with DFKI (German Center of Artificial Intelligence) we investigated the transformational impact of Generative AI in the area of software development for the business of our clients. Together with Alassane Ndiaye, Gerd Herzog (DFKI, German Center of Artificial Intelligence), Xhorxhina Taraj, Bostjan Robeznik, Dr. Christian Gerber Gerber and myself (Accenture) we wrote the paper "Generative AI in Software Engineering: Transforming the Software Development Process". In this paper, we assess the current and expected impact of generative AI (GenAI) on software engineering (SE), in particular on initial requirements gathering and design, implementation (including code generation, suggestion, and explanation), as well as from integration and testing to continuous deployment and support (code review, repair, and maintenance). The collaboration between human software engineer and AI tools is at the core of effective integration of GenAI into SE. AI can take over formalized, technical, repetitive, error-prone, and effort-intensive tasks. Human developers should focus on well-defined problem formulations, a creative outlining of the solution structure, its orchestration and supervision as well as a critical supervision and validation of AI-generated suggestions. In a holistic awareness of the different phases of SE, LLMs can fully unlock the impact of GenAI on productivity, efficiency, and innovation. Future challenges comprise questions of intellectual property and privacy of code and data, validation of AI suggestions (transparency and explainability of AI suggestions, availability of data sets), and the constant adaptation to the rapidly evolving field of AI. Paper link: https://lnkd.in/dMq6-sNB Press release: https://lnkd.in/dQm6YJSC Igor Aranovsky, Ian Kahan, Felix Klemm, Dr. Matthias Ziegler, Laetitia Cailleteau, Tobias Regenfuss, Arnab Chakraborty, Dr. Jochen Malinowski, Florian Schaudel, Nikhil Rajagopalan, Christopher Twardokus, Kathrin Schwan, Lukas Kölbl, Julia Staab, Thomas Seidel, Dr. Galia Diez, Bogdan Sacaleanu, Jonathan Abend, Arnaud Dolignon

72

7 Kommentare

Today SAP and NVIDIA's partnership goes further than ever in generative AI space!! Great announcements today from GTC, NVIDIA’s biggest conference of the year. Together, SAP and NVIDIA are redefining how enterprises use and benefit from AI—bringing automation, intelligence, and efficiency to every facet of business operations. 🚀 SAP is tapping Llama Nemotron models to advance SAP Business AI solutions and Joule, the AI copilot from SAP. Additionally, it is using NVIDIA NIM and NVIDIA NeMo™ microservices to promote increased code completion accuracy for SAP ABAP programming language models. “We are collaborating with NVIDIA to integrate Llama Nemotron reasoning models into Joule to enhance our AI agents, making them more intuitive, accurate and cost effective,” said Dr. Walter Sun, global head of AI at SAP. “These advanced reasoning models will refine and rewrite user queries, enabling our AI to better understand inquiries and deliver smarter, more efficient AI-powered experiences that drive business innovation.” #SAP #NVIDIA #GenerativeAI #Joule #JouleAgents #SAPBusinessAI #SAPBTP #Nemotron #ABAP #SAPdevelopers Jesper Schleimann David Pontoppidan Kevin Carlyon Luca Bucci Harish Kintali Mikkel Storm Møller Óttar Ómarsson Anthony Laursen Mathilde Sterndorff Cedric SIME Therese Elmhoff

35

🚀 We're thrilled to announce the arrival of the GPT-4.1 model series on Azure OpenAI Service: ✅ GPT-4.1 ✅ GPT-4.1 Mini ✅ GPT-4.1 Nano These powerful models deliver enhanced performance in coding, instruction following, and long-context understanding — empowering developers to build even more intelligent and scalable applications. 🔗 Learn more: https://lnkd.in/gGzgQgPE #AzureAI #AzureOpenAI #GPT #GenerativeAI #AzureFoundry #MVPBuzz #developers #followoz

23

🗓️ Today marks 1 year since the first version of #semanticlinklabs was released! In that time we have grown from 167 functions to over 450. A year ago, pretty much all of the functionality was focused on semantic models. Now, while there are even more capabilities around semantic models, we have further expanded the library to include other areas of #microsoftfabric including reports, lakehouses, warehouses, capacities, workspaces and much more! At this point in time, no single tool comes even close to matching the full capabilities of semantic link + semantic link labs. Just take a look at the high level feature list (https://lnkd.in/dwiMBfnh): 🔥Model Best Practice Analyzer 🔥View & edit semantic model metadata via the Tabular Object Model (#TOM) 🔥Vertipaq Analyzer (and creating a .vpax file) 🔥Format DAX 🔥Report Best Practice Analyzer 🔥View & edit report metadata via the ReportWrapper 🔥Run DAX, DMV, XMLA, or TMSL against a semantic model 🔥Run traces against semantic models 🔥Easy access to Power BI, Fabric, Azure, Microsoft Graph APIs 🔥Query Workspace Monitoring logs 🔥Access to AI (i.e. translations) and other python libraries 📌 Semantic Link Labs: https://lnkd.in/dAkqtNTc 💪 Bottom line - if you are using #powerbi or #microsoftfabric, there is something here which can help make you and your organization more productive, allowing you to focus on higher-level areas which truly need human attention. 🙏 A huge thanks to all the folks who have collaborated on this - both internally at #microsoft as well as within the Microsoft Fabric and Power BI community. On this note, many of the articles and videos which were published on semantic link labs have been consolidated here on the Wiki: https://lnkd.in/dtJ8anbc. 🗣️If semantic link labs has influenced the way you and your organization work - please share the feedback. We'd love to hear from you! 💡If you have an idea for something which you feel should be included in semantic link labs, let me know (https://lnkd.in/d6p28srA)! #semanticlink #elegantbi #powerbi #python #jupyternotebook

251

28 Kommentare

Be honest: does your DS team have a mission‑critical model in prod? Most don’t. Fraunhofer 2021 report says only 1 in 4 data projects cross the line. So if your Data Science team stuck in PoC limbo and misunderstood, you're a not alone. But to navigate next budget cut, you need to act now! I’m about to share my framework with you. 👉 Why should you trust me? I implemented this framework successfully as engineering manager (see profile) and I talk regularly to dozens of Data-Science Managers who do the same using my open-source backend for KeplerGL. 👉 Why Data‑Science Teams Struggle to Stay Relevant? → Engineering hand-off: You finish the model, then wait weeks for Backend. → Fuzzy business/product asks: they don’t understand/trust the models. A confused mind always says “no.” → Legal friction: every new launch means privacy checks and security sign‑offs. Nobody wants to take risks. But you can navigate it in simple (but not easy) 3 steps: 👉 The 3‑Step “Stay‑Relevant” Framework. STEP 1: Serve an Internal Power‑User → Pick one decision‑maker with real authority: Ops, Finance, Sales. → Serving internal users cuts the product + legal dependecy → you own the roadmap. → When your model steers high‑stakes calls, the team becomes mission‑critical. STEP 2: Ship a Data Product, not a Real‑Time Model. → Ask: “Which decision still runs on gut?” Build the dataset that fixes it. → Company Data Platform is your production, batch notebooks + Papermill / GitHub Actions are enough → cut dependency to engineering → Some model in use beats a perfect model ignored. If they ask for better accuracy, you’ve already won. STEP 3: Let Stakeholders Pull the Insight Themselves. → Wrap the table in a self‑serve view: Metabase, Streamlit, or a KeplerGL/Dekart map. → When stakeholder discover the answer themselves, it becomes their idea → ownership → trust → action. → Pro tip: Maps beat Dashboards on context. Maps are great communication tool → deliver maps when you can. 🔑 Summary: Internal power‑user → data product → self‑serve delivery. Get that loop right and your DS team shifts from cost center to can’t‑live‑without, lifting the whole org in the process. Win‑win 🤝 👉 IMPORTANT: Know someone who should see this? Do them a favor: tag them in comments below. #DataScience #MLOps #AnalyticsLeadership #DataProducts #BudgetCuts

19

3 Kommentare

The future of #Windows is #Cloud & #AI. Yesterday we announced #Researcher with Computer Use in #Microsoft365Copilot. A major leap toward more autonomous AI that works for you. Now with #ComputerUse, #Researcher goes beyond reasoning and researching. It can now act on your behalf using a secure virtual computer to navigate #public, #gated, and #interactive web content. This powerful extension of Researcher—combined with the unique ability to connect to your work data—unlocks smarter research, deeper insights, and more comprehensive reports. And #WindowsCloud is providing the foundation so Researcher and other #ComputerUseAgents to perfom these actions. Read more here: https://lnkd.in/d3Zxbsyu Share your favourite research prompt in the comments? #UnlockedPotential #Windows365 #FutureOfWork

39