Enhance Your Machine Learning Workflow with DeepSeek in SageMaker Studio   https://lnkd.in/e3JT9rJP   Machine learning teams and data scientists working in AWS environments can supercharge their productivity by integrating DeepSeek with SageMaker Studio. This powerful combination transforms how you build, train, and deploy ML models by streamlining complex workflows and automating repetitive tasks.

🚀 ModelAdvisor v1.0 Launch 🎥 Introducing ModelAdvisor v1.0 an intelligent benchmark framework for AI model selection Selecting the right LLM isn’t just about performance it’s about cost, latency, accuracy, and safety. That’s why I built ModelAdvisor, a framework that helps Technical Product Managers and AI Solutions Engineers make data-driven decisions when choosing AI models. 🔹 Version 1.0 focuses on Amazon Bedrock, benchmarking models like Claude and Nova. It is also connected with AWS CloudWatch. Run prompts, measure latency, compute cost per inference, apply guardrails, and export results all in one workflow. 💡 The result: ModelAdvisor can reduce by up to 60% the time it takes to find the ideal model for your project — helping teams move from guessing to knowing. How it helps ⚙️ Runs identical prompts across providers 📊 Measures cost, tokens, and latency 🧠 Evaluates semantic accuracy 🛡️ Enforces safety & PII guardrails ☁️ Logs results to AWS CloudWatch We’re planning to expand beyond Bedrock to support more providers soon. 💬 What else would you consider important when evaluating an AI model? (Factuality? Tool usage? Context handling?) Share your ideas below they’ll help shape the next version! 🎥 Watch the demo below 💻 GitHub repo with full code and setup guide in the comments. #AI #LLM #AmazonBedrock #Benchmark #AIEngineering #ModelAdvisor #GenAI #ProductManagement #MachineLearning #AIInfra #AIEvaluation

I've been working on a new data pipeline for a project, and the latest features in AWS Glue 5.0 are so amazing. The integrated generative AI capabilities have been a huge time saver for me. Specifically, I've been using the new AI-powered troubleshooting feature. Instead of spending hours manually sifting through logs to pinpoint a root cause, AWS Glue analyzes job metadata and execution logs to provide an immediate diagnosis and actionable recommendations. What used to be a long debugging session has been reduced from hours to minutes. This isn't just about faster fixes, it's about focusing on higher value tasks and building more resilient data products. #DataEngineering #AWS #AWSCloud #AWSGlue #GenerativeAI #ETL #CloudComputing

• New StartSession API enables secure SSH-over-SSM tunnels from VS Code to SageMaker Studio spaces, ending the “SSH workaround tax” and auto-reconnecting to protect work. • Autoscaling observability collectors unify GPU/network/storage metrics on HyperPod, detect grey failures, and cut root-cause time from days to minutes with a single dashboard. • HyperPod now supports deploying models on the same clusters used for training, unifying train-and-serve to boost utilization and simplify operations. • New Kubernetes HyperPod training operator improves fault recovery, restarting only affected resources and monitoring stalled batches/NaNs with YAML-configurable policies. 🔔 Follow us for daily AI updates! 📘 Facebook: https://lnkd.in/gxDt7PJa 📸 Instagram: https://lnkd.in/gmYfWDbF #AWS #AmazonSageMaker #GenerativeAI #MLOps #AIGenerated #CreatedWithAI

🚀 Excited to share my latest project: AI-Powered Claims Similarity Search! This solution empowers claims professionals to: • Instantly find similar historical claims using semantic search • Accelerate claim review and decision-making • Reduce manual research and improve accuracy 🔍 How it works: - Uses Amazon Bedrock embeddings and S3 Vector Search to compare new claims against thousands of past cases - Returns the top matches with similarity scores and visual indicators - Enables fast, data-driven insights for adjusters and analysts 💡 Built with: - .NET 9, Semantic Kernel, AWS Bedrock, S3 Vectors - Modern web UI for easy access and visualization Whether you’re handling workers’ compensation, property, or casualty claims, this tool streamlines the process and helps teams make better decisions, faster. #AI #MachineLearning #AWS #VectorSearch #InsureTech #SoftwareEngineering

Last week, I co-hosted the AWS User Group Strasbourg meetup with Charles Rapp (Amazon Web Services (AWS)) and Hakim Rachidi (Scafe). Our talk “Master Generative and Agentic AI with AWS” focused on moving beyond demos, showing how to build production-ready agentic AI. 🚀 Amazon Bedrock AgentCore delivers the runtime foundation for scalable AI agents, with built-in memory, observability, identity, and secure tool orchestration, key to running AI in enterprise environments. 🧑💻 The Strands Agent SDK offers a code-first, lightweight way to develop and deploy agents quickly, giving developers flexibility to experiment while staying on a path to production. 🧩 Combined, they enable enterprises to go from prototype to production with governance, reliability, and scale. Right after the presentation, we ran a hands-on workshop where participants could experiment with these tools, turning theory into practice and seeing how agentic AI can be built for real-world use cases. At TrackIt, we’re already applying these frameworks in client projects, helping organizations adopt generative and agentic AI that delivers measurable business value, not just proofs of concept. Thank you to everyone who joined us, and to Amazon Web Services (AWS) and Scafe for making this session possible. The momentum around production-grade AI is real, and we’re just getting started. #AWS #Bedrock #AgentCore #Strands #GenerativeAI #AgenticAI #TrackIt

Most GenAI pilots never make it to production. Why? Because they lack the foundation for scalability, governance, and enterprise alignment. In this week’s TechTips, we’re sharing 5 proven strategies to help you build business-ready GenAI solutions on AWS. From exploring the world of Amazon Bedrock, to mastering prompt engineering, we tell you all you need to know about AWS GenAI solutions. Looking to put theory into practice? Our instructor-led course, Generative AI Essentials on AWS, equips you with the skills to turn GenAI concepts into enterprise-grade applications. In just 8 hours, the course will get you hands-on with AWS AI/ML solutions and show you how to lead and execute an end-to-end GenAI project. Register today at 50% off! Offer valid for a limited time. Click here to register: https://lnkd.in/gs3QGY6x #AWSTraining #GenAI #AWSBedrock #NetComLearning

🚀 How fast can you really take an AI/ML platform from idea to production? Sepi Mohammadi will show us! At apotea.se, his team built a distributed, serverless AI/ML platform in just eight weeks, successfully launching three products without the usual headaches. ⚡ In this session, Sepi will walk through the maturity plan, tools, and strategies that made it possible—from SageMaker and Databricks to lightweight custom setups designed for speed and scale. ✨ A practical look at how businesses can deliver AI/ML platforms that are both fast and scalable—without breaking the budget or the team. #awscommunity #aws #portugal #cloudcomputing

𝗔𝗜 𝗠𝗶𝗰𝗿𝗼𝘀𝗲𝗿𝘃𝗶𝗰𝗲𝘀 𝗗𝗲𝘀𝗶𝗴𝗻 𝗣𝗮𝘁𝘁𝗲𝗿𝗻 - 𝘀𝗵𝗶𝗽 𝗮𝗴𝗲𝗻𝘁𝘀 𝘁𝗵𝗮𝘁 𝗱𝗼𝗻’𝘁 𝗺𝗲𝗹𝘁 𝘆𝗼𝘂𝗿 𝘀𝘁𝗮𝗰𝗸 Most AI failures aren’t about the model. They’re about scope, side-effects, and ops. What works is an AI microservices approach: small, single-purpose agents + safe tools + strong guardrails. 𝗣𝗮𝘁𝘁𝗲𝗿𝗻 (𝗸𝗲𝗲𝗽 𝗶𝘁 𝗯𝗼𝗿𝗶𝗻𝗴, 𝗸𝗲𝗲𝗽 𝗶𝘁 𝗹𝗶𝘃𝗲):  1. One agent = one outcome (clear KPI: goal completion or $/resolved task).  2. Tool adapters with contracts (idempotency, SAGA/comp, DLQ).  3. State & memory separated (KV/vector) with TTL & PII minimization.  4. Workflow via queues/orchestrators; no hidden side-effects in prompts.  5. Eval & observability: cost, p95 latency, override rate, success. 𝗚𝘂𝗮𝗿𝗱𝗿𝗮𝗶𝗹𝘀:  • Policy-as-code in CI/CD and runtime (access, lineage, audit).  • Budget caps + auto-suspend per agent.  • Portability: choose Microsoft/OpenAI/AWS/Google; run on Databricks/Snowflake when it’s cheaper/faster. At ARCHISURANCE, we turn this into golden paths so teams move from clever demos to dependable systems fast. 𝗤𝘂𝗲𝘀𝘁𝗶𝗼𝗻: If you had to pick one KPI for your next agent, what is it - $ per task, goal completion, or override rate? #AIArchitecture #AIAgents #EnterpriseArchitecture #DataArchitecture #DataEngineering #MicrosoftAI #OpenAI #AWS #GoogleAI #Databricks #Snowflake #CTO #TechLeadership #Archisurance