Sami Ahma-Aho

99% of people are just using LLMs. The top 1%? They’re building them. 💻🧠 Sebastian Raschka, PhD just launched something big: Master and Build Large Language Models → 👉 https://lnkd.in/dzKMZi_p I’ve read all his books. I’ve gone through every course. This one is next-level. No shortcuts. No copy-paste notebooks. You build the architecture yourself. From scratch. 💥 What you’ll learn: → Build a transformer, layer by layer → Train and fine-tune real LLMs → Deploy them into production → Understand tokenization, attention, gradients—the whole thing This is not for tourists. It’s for builders who want to understand how LLMs actually work. Line by line. Block by block. If you’re tired of surface-level AI noise, here’s the signal. This course gives you the technical foundation everyone else is skipping. I’m diving in. Let’s go deeper. 🔧📘

101

8 kommenttia

💡 85% of AI projects fail—not because of weak models, but because of bad data. Before jumping into AI, ask yourself: 1️⃣ Is your data accessible across teams? 2️⃣ Is it trustworthy and governed? 3️⃣ Is it relevant to real business outcomes? 🚫 Without strong data foundations, even the best AI becomes noise. 📥 I’ve shared a practical framework to evaluate your org’s AI readiness. #AIReady #DataStrategy #AIGovernance #DigitalTransformation #EnterpriseAI #AILeadership #AIAdoption #DataQuality

273

40 kommenttia

🧠 “The new #Edge #Copilot Just Analyzed #NovoNordisk for Me — Is It a Buy? Here’s What I Found” I put the new tab-aware Copilot mode in Microsoft Edge to the test and asked a big question: Is Novo Nordisk a solid long-term buy? 🚀 Instead of jumping across tabs, I let Copilot handle the heavy lifting—scanning Reddit sentiment, investor commentary, earnings reports, and the company’s R&D pipeline across multiple open tabs. Fast, focused, and surprisingly smart. 💡 What Copilot surfaced: Major stock dip after slashed 2025 guidance Strong diabetes and obesity drug pipeline still intact Valuation hit multi-year lows—serious long-term potential? 🎥 I recorded the whole thing to show how Copilot works in action. Check out the video in the comments—this could change how you research and invest.... or maybe how you find the recipe for the perfect "Drømmekage" - totally up to you 🍰 👉 Learn more about the new Edge Copilot mode on the official Microsoft Edge blog https://lnkd.in/dqXrzwUy #NovoNordisk #MicrosoftEdge #Copilot #AIinvesting #StockAnalysis #PharmaTech #FutureTools Brian Guldborg Mette Louise Kaagaard Henri Schulte Jens Christian Schrøder Ali Reza Farahnak Sonika Kapil Peter Lunding Smith Rasmus Jensen Mustafa Suleyman

46

I still see many teams comparing AI tools by features, then re-architecting six weeks later. So I mapped 12 real enterprise scenarios across LangGraph, LangChain, n8n, and AutoGen to make the choice obvious. Easy to understand example: Example: New employee onboarding (Day 0 → Day 1) Goal: Get laptop + accounts + access live in 24 hours, with approvals and audit. LangGraph: Model as a clear flow: HR webhook → verify docs → create checklist → request laptop → pause for manager approval (licenses/cost) → provision apps → confirm → if any step fails, resume/rewind from checkpoints (“time travel”) and retry. Great for guardrails and resumable steps. LangChain: Use as the LLM brain to read offer/role and generate: app list, access scopes, welcome pack, FAQs. Pair with another system to actually provision. n8n: Best for the glue: receive HRIS event → create Okta/Google Workspace users → open IT ticket → send Slack “Welcome” → calendar invites → approvals via approval patterns (Slack send-and-wait, forms, webhooks) → log everything to a sheet/DB. Low-code, fast. AutoGen: Planner + Tool-User agents decide the sequence and call APIs; add a supervisor to keep them on track. Useful if onboarding varies a lot by role—add strict stop conditions before any real changes. Routing rules Governed, recoverable steps → LangGraph Content/logic generation (who needs what, why) → LangChain Integrations, webhooks, approvals → n8n Flexible agent planning (lots of variations) → AutoGen Please share your experiences too . #AI #AgenticAI #LangGraph #LangChain #n8n #AutoGen #RAG #LLMOps #AIOps #EnterpriseAI P.S. All views are personal

46

15 kommenttia

OCR document parsing metrics create a fundamental problem: they punish generative models for being "different," even when they're actually right. A model can extract every piece of information perfectly, but if it formats the output slightly differently than expected, it gets penalized by rigid evaluation criteria. This mismatch between technical accuracy and practical utility led us to develop SCORE, a new semantic evaluation framework for generative document parsing. Rather than fixating on exact format matching, SCORE evaluates what truly matters: whether a vision-language model actually understood and preserved the document's content, structure and meaning. You can explore the full methodology in our paper: <https://lnkd.in/gSDe_3Pk> This philosophy of prioritizing substance over form extends to how we think about parsing solutions more broadly. The best outcomes come from flexibility: choosing approaches that align with your specific data, workflows, and business requirements. That's precisely why Unstructured doesn't lock you into a single parser or model. Our platform provides multiple parsing strategies and seamlessly integrates with leading vision-language models—from Claude to GPT-4o to Gemini—with new options continuously added as the field advances. Every solution we offer is rigorously benchmarked against real-world scenarios. As new models and techniques emerge, we evaluate them, optimize their performance, and make them available to you. This means you're never constrained by today's capabilities and continuously benefit from tomorrow's advancements without the friction of constant migration or integration work.

58

2 kommenttia

🚩 Agentic AI: Top 5 Myths vs Facts   Agentic AI is just a chatbot… right?   Not even close.   Agentic AI isn’t just about conversation, it’s about planning, reasoning, taking action, and using tools.   But with so much noise out there, a lot of myths still make the rounds.   Top 5 Myths About Agentic AI from “you need AGI” to “it’s too early for enterprise use.” These common misconceptions are slowing down adoption or stopping teams before they start.   The reality?   Agentic AI is already in use across industries powering email triage, underwriting assistants, claims workflows, and more.   #AgenticAI #AIagents #GenerativeAI #LLMs #Innovation

49

3 kommenttia

Building Agentic solutions that rely on a Language Model are wrappers offer no value add. An Agentic Architecture will have to be designed and engineered! Devising an architecture requires both an in depth understanding of the business goals and a deeper understanding of AI tools and the landscape. Making it both engineering and art. 1/ A good agentic design should be model agnostic and use case specific. Knowing to leverage Small Language models in orchestrating workflows is a breakthrough in managing costs, and latency. The NVIDIA paper Small Language Models are the Future of Agentic AI is a dialectical account of this -  https://lnkd.in/ghVhu9ur 2/ If you are thinking of knowledge management to models, think Context Engineering and not RAG. Actively & continuously structuring the inputs into an LLM’s context window is key to unlocking accuracy and minimizing hallucinations. This post by SRK summarises everything about Context Engineering - https://lnkd.in/g4ywnGdn 3/ Knowing how LLMs function involves making sense of the LLM’s Semantic layer ( can be called vibing !?). Being able to map business, technical, context data into a semantic structure that the LLM gets is a very practical and lived experience. Read this post by Zaher Alhaj https://lnkd.in/gkQhAn6A 4/ Planning of LLMs – Agentic Layers requires to plan at each step using inference time reasoning. Getting to know the right inference time reasoning techniques to use such that cost per task, latency, steps per task, context adherence are all optimized for is a skill. Knowing about all the planning techniques in a single place is here - https://lnkd.in/g4ZgAakD 5/ Agentic applications go from a POC to Production only using Evals. Evals enable to devise a structured approach to Test, Verify, Iterate to build Agentic applications. Evals is turning out to be the most fundamental value add for production ready systems. Hamel H. has been harking about it for more than a year see his FAQs for Evals here - https://lnkd.in/gPjMvNTt Beyond the hype that AI solutions can do magic and are transforming in breakneck speed. For Agentic AI to become a ‘normal technology’ is a marathon – steady in its own pace. Maturity in technology is the virtuous cycle Innovation -> Adoption -> Diffusion. https://lnkd.in/giaRmtPQ Lastly, an area that I personally believe will re define agent development is Post Training Language Models [PoLM]. There is a lot to learn and experiment here. Here is an encyclopaedia of it - https://lnkd.in/gk_TZ3B4 Keep it going AI PMs, AI Architects and AI Engineers !

30

1 kommentti