About
27+ years of full-stack engineering and AI innovation, turning complex challenges into…
Services
Articles by Edmund
-
Designing Scalable Enterprise Agentic AI: A 2026 Reference Architecture on Microsoft Foundry
Designing Scalable Enterprise Agentic AI: A 2026 Reference Architecture on Microsoft Foundry
As an AI Architect based in the Seattle area (Bellevue vibes included), I've spent the past year diving deep into…
Activity
7K followers
-
Edmund T. posted thisAI Hallucinations Aren’t a Bug - They’re a Byproduct of How We Build AI There’s a common misconception that modern AI systems “know” facts. They don’t. Large Language Models (LLMs) learn statistical patterns from data, not truth. When that data includes noise, bias, or misinformation (as much of the internet does), the model can produce outputs that are: Fluent Confident …and sometimes wrong This is what we call hallucination. Why Hallucinations Happen? At a technical level, LLMs are optimized to predict the most likely next token, and not the most verified fact. That leads to: False positives (incorrect but plausible answers) Fabricated details (e.g., citations, facts, or events that don’t exist) Overgeneralization from biased or incomplete data Even with fine-tuning and human feedback, this behavior cannot be fully eliminated - only controlled and mitigated. Reframing the Problem Instead of asking: “How do we eliminate hallucinations?” A better question is: “How do we design AI systems that know when they might be wrong and verify before answering?” Practical Solutions (That Actually Work) 1) Retrieval-Augmented Generation (RAG) Ground responses in trusted, real-time data sources instead of relying only on model memory. Connect LLMs to curated knowledge bases, Force citation-backed responses, Reduce fabrication risk significantly 2) Confidence & Uncertainty Modeling Make AI express how sure it is. Probability scoring Calibration techniques Threshold-based response filtering If confidence is low → defer, ask clarifying questions, or escalate. 3) Multi-Model Verification (Ensemble Systems) Don’t trust a single model. Cross-check outputs across multiple models Use consensus or voting mechanisms Flag disagreements for review 4) Tool-Using Agents Let AI verify before responding. Call search APIs Query databases Execute code for validation Shift from “generate answer” to “generate + validate answer” 5) Human-in-the-Loop (Critical Systems) For high-stakes domains (healthcare, finance, national security): AI proposes ---> human verifies ---> system learns Continuous feedback loop improves reliability 6) Domain-Specific Fine-Tuning General models struggle with specialized truth. Train on high-quality, domain-curated datasets Apply strict data governance The Strategic Insight Hallucination is not just a technical issue - it’s a system design problem. The future of reliable AI will not rely on: bigger models alone …but on: well-architected systems that combine models, tools, data, and verification layers Bottom Line AI should not be treated as a source of truth, but as a: reasoning engine that must be grounded, validated, and monitored Organizations that understand this early will build: More trustworthy systems Safer AI applications Real competitive advantage
-
Edmund T. shared thisBecoming an agentic AI engineer is not a prompt engineering skill. It is the ability to design systems that combine: reasoning (LLMs) control (state machines) tools (APIs & services) memory (context persistence) evaluation (measurable correctness) You are not building chatbots. You are building autonomous software systems with decision loops.
-
Edmund T. shared thisThe Real Path to Becoming an Agentic AI Engineer Agentic AI engineering is not about using LLMs. It is about building autonomous, stateful, tool-driven systems that operate reliably under uncertainty. The progression moves from deterministic systems to controlled autonomy to scalable agent architectures. Phase 1 - Foundations (Deterministic Systems First) Goal: Understand LLMs as probabilistic components inside software. Core concepts Tokens, context windows, latency, cost behavior Prompting: instruction, few-shot, structured outputs (JSON/schema) REST APIs, authentication, retries, rate limits Async programming and error handling Output validation and schema enforcement Key mindset LLMs are external unreliable services, not reasoning engines. Build LLM wrapper with retries and logging Prompt to structured JSON extractor API integration service with validation Phase 2 - Agent Core (Control Loop) Goal: Build the fundamental execution loop. Agent loop Observe Think Act Evaluate Update State Repeat Core concepts Tool function calling Explicit state management Memory systems short term working context long term vector store episodic history ReAct reasoning pattern Tool registry design Key mindset An agent is a state machine with reasoning, not a prompt. Build ReAct agent with trace logs Tool system for search API calculator Persistent state object Basic memory store Phase 3 - Architecture (Multi Agent Systems) Goal: Move from single agents to systems. Core concepts Retrieval Augmented Generation Task decomposition and planning Multi agent patterns orchestrator and workers planner executor critic roles Failure handling and retries Key mindset Architecture is about reliability and failure control. Build RAG agent with grounding Planner Executor Validator pipeline Async task queue Fallback strategies Phase 4 - Safety and Evaluation Goal: Make autonomy measurable and safe. Core concepts Guardrails and schema validation Human in the loop checkpoints Metrics success rate hallucination rate tool accuracy cost latency Observability with step level logs Key mindset If it cannot be measured, it cannot be trusted. Build Evaluation harness with test cases Prompt and version tracking Feedback loop system Safety filter layer Phase 5 - Production and Scale Goal: Turn agents into production infrastructure. Core concepts Integration with APIs CRMs databases Deployment with containers serverless workers Observability logs traces metrics Security secrets and role based tool access Key mindset Scaling is not more agents it is fewer failures per execution. Build Production agent API service Queue based execution system Monitoring dashboard for cost steps errors Secure tool gateway
-
Edmund T. shared thisVector RAG vs. Vectorless RAG: Moving Beyond Traditional Vector Databases
-
Edmund T. posted thisVector RAG vs. Vectorless RAG: Moving Beyond Traditional Vector Databases? For the past year, vector databases have served as the default foundation for Retrieval-Augmented Generation (RAG). But as Large Language Models continue to advance, a fundamental question is arising: What if your AI system could truly reason through information instead of simply searching it? Enter Vectorless RAG (also known as Reasoning-based Retrieval or PageIndex). If you're designing next-generation AI applications, understanding these two approaches is crucial. Here's a clear side-by-side comparison: Vector RAG: Machine-Like Search The classic pipeline: Documents are split into chunks, converted into vector embeddings, and retrieved via similarity search. Strengths: Extremely fast and highly scalable Excellent for general-purpose, high-volume retrieval Limitations: Relevance is only approximate Chunking often breaks important context Similarity scores are largely opaque ("black box") Vectorless RAG: Human-Like Reasoning Instead of fragmenting content, this method builds a structured, hierarchical index of the full document. The LLM then intelligently navigates this structure through multi-step reasoning to locate the most relevant sections. Strengths: Preserves complete document context No vector database required Fully transparent and traceable reasoning paths Dramatically better accuracy on complex, nuanced queries Limitations: Higher computational cost Slower than pure vector search Which Should You Choose? Choose Vector RAG when speed and scale matter most - ideal for customer support bots, broad document search, and knowledge bases. Choose Vectorless RAG when precision, explainability, and reliability are critical - perfect for legal contract analysis, medical record review, financial audits, and other high-stakes enterprise use cases. As LLMs become faster and more efficient, reasoning-based retrieval approaches like Vectorless RAG are poised to become the new gold standard for sophisticated AI systems. What are your thoughts? Have you tried vectorless or hierarchical retrieval methods yet? Share your experiences below! 👇
-
Edmund T. liked thisEdmund T. liked thisAre we still explaining AI to business leaders the wrong way? Most conversations about AI start with models, tools, or buzzwords. But business leaders do not care about how AI works. They care about what it changes for the business. This visual gets it right. At the core, AI and ML turn data into decisions like forecasting, pricing, fraud detection, and churn prediction. On top of that, neural networks detect patterns humans cannot like speech, images, and sentiment. GenAI adds speed by creating content, code, and answers at scale. AI agents go further by executing tasks on their own like resolving tickets or processing expenses. Agentic AI sits at the top, automating entire workflows across systems from onboarding to financial close. The shift is important. AI is moving from insights to actions. From assistance to autonomy. From tools to outcomes. If you want business buy-in, stop explaining AI by technology layers. Explain it by the work it removes, the time it saves, and the decisions it improves. Credits: John Wernfeldt
-
Edmund T. liked thisEdmund T. liked thisClaude AI Cheat Sheet 𝗕𝗲𝗰𝗼𝗺𝗲 𝗯𝗲𝘁𝘁𝗲𝗿 𝗮𝘁 𝗔𝗜 𝗶𝗻 𝗷𝘂𝘀𝘁 𝟭 𝗺𝗶𝗻𝘂𝘁𝗲 𝗮 𝗱𝗮𝘆. 𝗚𝗲𝘁 𝘁𝗵𝗲 𝗔𝗜 𝗻𝗲𝘄𝘀𝗹𝗲𝘁𝘁𝗲𝗿 𝘀𝗺𝗮𝗿𝘁 𝗹𝗲𝗮𝗱𝗲𝗿𝘀 𝗿𝗲𝗮𝗱. 𝗦𝗶𝗴𝗻 𝘂𝗽 𝗳𝗿𝗲𝗲 𝗻𝗼𝘄 → aiforleaders.com Original post: __________ 1. Claude Models Use the right model for the job. • Opus 4.5 → Hard reasoning, research, complex tasks • Sonnet 4.5 → Daily writing, analysis, editing (best default) • Haiku 4.5 → Fast, cheap tasks and quick prompts All models support 200K context, which means you can feed large documents and projects. 2. Prompting Techniques The quality of your output depends on the structure of your prompt. Some of the most effective techniques: • Role playing • Chained instructions • Step-by-step prompting • Adding examples • Tree of thought reasoning • Style-based instructions The best combo usually is: Role + Examples + Step by Step. 3. Role → Task → Format Framework One of the simplest ways to improve prompts. Example structure: Act as [Role] Perform [Task] Output in [Format] Example: Act as a marketing expert Create a content strategy Output in a table or bullet points 4. Prompt Learning Methods Different prompt styles produce different outputs. • Open ended → broad exploration • Multiple choice → force clear decisions • Fill in the blank → structured responses • Comparative prompts → X vs Y analysis • Scenario prompts → role based thinking • Feedback prompts → review and improve content 5. Prompt Templates You can dramatically improve results using structured prompting. Three core styles: • Zero shot → no examples • One shot → one example provided • Few shot → multiple examples More examples usually means better outputs. 6. Projects Projects turn Claude into a knowledge workspace. You can: • Upload files as knowledge • Organize chats by topic • Add custom instructions • Share with teams • Maintain long context across work 7. Artifacts Artifacts allow Claude to generate interactive outputs like: • Code • Documents • Visualizations • HTML or Markdown apps You can read, edit, and run them directly inside the chat. 8. MCP + Connectors MCP (Model Context Protocol) connects Claude to external tools. Examples: • Google Drive • Gmail • Slack • GitHub • Figma • Asana • Databases 9. Claude Code Claude can also act as a coding agent inside the terminal. It can: • Read entire codebases • Write and test code • Run commands • Integrate with Git • Deploy projects 10. Reusable Skills + Hooks Claude supports reusable markdown instructions called Skills. Plus automation hooks like: • PreToolUse • PostToolUse • Stop • SubagentStop These help control workflows and outputs. Prompt Starters Some prompts work almost everywhere: • “Act as [role] and perform [task].” • “Explain this like I am 10” • “Compare X vs Y in a table.” • “Find problems in this document.” • “Create a step-by-step plan for [goal].” • “Summarize in 3 bullet points.” Credit to Shah Riyed Arifen. Follow him for more.
-
Edmund T. liked thisEdmund T. liked this→ 𝐀𝐫𝐞 𝐲𝐨𝐮 𝐬𝐭𝐢𝐥𝐥 𝐭𝐫𝐞𝐚𝐭𝐢𝐧𝐠 𝐚𝐥𝐥 𝐝𝐚𝐭𝐚𝐛𝐚𝐬𝐞𝐬 𝐭𝐡𝐞 𝐬𝐚𝐦𝐞 𝐰𝐚𝐲 The world of data is evolving. Traditional relational databases are powerful, but modern applications demand more flexibility, speed, and scalability. That’s where NoSQL databases come in. 𝐇𝐞𝐫𝐞 𝐚𝐫𝐞 𝐭𝐡𝐞 𝐭𝐨𝐩 9 𝐲𝐨𝐮 𝐬𝐡𝐨𝐮𝐥𝐝 𝐤𝐧𝐨𝐰: • MongoDB – Document-oriented, highly scalable, great for modern web apps. • Cassandra – Distributed, fault-tolerant, ideal for handling massive amounts of data. • Redis – In-memory key-value store, perfect for caching and real-time analytics. • Couchbase – Combines key-value store and document database, optimized for high performance. • Neo4j – Graph database, excels at mapping complex relationships. • Amazon DynamoDB – Fully managed, serverless NoSQL database for ultra-fast applications. • Apache HBase – Column-oriented, designed for big data on top of Hadoop. • Elasticsearch – Search engine database, enables powerful real-time search and analytics. • CouchDB – Document database, reliable replication and offline-first capabilities. Choosing the right database can make or break your application’s performance. Understand your data model, query patterns, and scalability needs before deciding. Follow Ashish Sahu for more insights
-
Edmund T. liked thisEdmund T. liked this12 Phrases That Kill Work Conversations (and what to say instead) You hear them every day. They seem innocent. Until you realize they destroy work relationships fast. 🚨 After treating 30,000 patients, I finally understood this. Time to flip these conversation killers into something better. Here are the worst ones, and how to turn them into trust-builders: "You don't understand" ❌ "Maybe I can explain better..." ✅ "I already told you..." ❌ "Let me explain this another way..." ✅ "You should have..." ❌ "Next time, let's try..." ✅ "That's not my problem" ❌ "Let's find who can help..." ✅ "You're being too sensitive" ❌ "This sounds important to you - tell me why" ✅ "I'm just being direct" ❌ "Here's my perspective..." ✅ "As per my previous email..." ❌ "To recap the key points..." ✅ "No offense, but..." ❌ "Can I share some thoughts?" ✅ "Do what you want..." ❌ "Here's what I recommend..." ✅ "Fine, have it your way" ❌ "Let's find middle ground..." ✅ "Just deal with it" ❌ "How can I support you?" ✅ "Yeah, but..." ❌ "I hear you, and..." ✅ The pattern? Toxic phrases shut people down. Safe phrases help them open up. Next time you're about to use one of these: Breathe. Choose words that build bridges, not walls. ____ P.S. These phrases help me build my business career. What's a phrase that helps you at work? 💬 ♻️ Follow Dr. Miro Bada and share this to help others 📌 Save this post for future reference! If you want a copy of my top 60+ infographics (free): 👉 Like, Repost, then Signup here: www.PeakProtocol.co
-
Edmund T. liked thisEdmund T. liked thisEvery developer using Claude Code makes the same mistake first. They jump straight into coding. No plan. No context file. No rules defined. No tests written first. Just: "Write me a function that does X." Then 2 hours of refactoring what Claude built without understanding what you actually needed. That’s not a Claude problem. That’s a workflow problem. Here’s the workflow that fixes it The prompt that changes everything: "Explain how you'd solve X. No code yet." Three words - No code yet - force Claude into planning mode. Review the plan. Refine it. THEN execute. Every hour spent planning saves 3 hours of refactoring. Every Claude Code power user knows this. Almost nobody does it on day one. The other 7 rules that compound on top: 📌 CLAUDE.md → your rules, every session, automatically 📌 TDD → failing test first, implementation second 📌 Subagents → parallel tasks, not sequential bottlenecks 📌 Git Worktrees → one feature, one environment, zero conflicts 📌 /compact → clean context, consistent quality 📌 /cost → visible spend, no surprise bills 📌 Skills → standardized workflows, consistent team output 8 rules. All learnable in one afternoon. All changing your output permanently. Which rule do you wish you'd learned on day one? Save this, the Claude Code workflow guide every developer needs before day two. ♻️ Repost for every dev who jumped straight into coding and paid for it in refactoring hours.
Experience
-
Earth's Element (Landscape Architects, Designers, Landscaping)
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
Education
Other similar profiles
Explore more posts
-
Lexi Reese
The last two weeks marked a *major shift* in the NOW of work. - 𝗪𝗵𝗮𝘁 𝗵𝗮𝗽𝗽𝗲𝗻𝗲𝗱: Claude Co-Work and the newest Codex agents moved from “assist” to “execute.” These systems can modify entire codebases, run full test suites, draft financial summaries, and reconcile data across tools. This isn’t autocomplete. It’s task ownership. It's the kind of work engineers, analysts, and associates bill for today. It's a lot of the work I do today, maybe you too! - 𝗣𝘂𝗯𝗹𝗶𝗰 𝗺𝗮𝗿𝗸𝗲𝘁𝘀: Public software is ~40% off 12-month highs. Investors are stress-testing companies like Salesforce, ServiceNow, Atlassian, and Adobe whose growth depends on adding seats. If AI trims even 10–20% of knowledge work, seat expansion slows. Services-heavy firms like Accenture feel it if automation compresses billable hours. This isn’t extinction. It’s margin compression. - 𝗣𝗿𝗶𝘃𝗮𝘁𝗲 𝗺𝗮𝗿𝗸𝗲𝘁𝘀: Capital is rotating toward AI-enabled services and outcome pricing. The pitch isn’t “more licenses per account.” It’s “2–3x throughput per employee,” 10–18% cycle-time reductions, and measurable revenue-per-rep or revenue-per-employee lift. Pricing is increasingly tied to EBIT impact — not seat count. That’s a structural shift in value capture and away from seat pricing b/c, well, there are less seats but also I (enterprise buyer) want to pay for value + AI services are basically promising margin improvement. - 𝗘𝗻𝘁𝗲𝗿𝗽𝗿𝗶𝘀𝗲 𝗹𝗲𝗮𝗱𝗲𝗿𝘀: If one AI agent can handle first-pass code review, draft client reports, or generate analysis, the org chart changes. The real question becomes: what jobs must humans uniquely do to build, sell, and support the product. And which layers are leverage? - 𝗔𝗹𝗹 𝗼𝗳 𝘂𝘀: Knowledge work is being repriced. “Leverage per employee” means one person, equipped with AI, can produce what previously required multiple people: more output, faster cycle times, higher revenue per head. But leverage cuts both ways. Expectations rise. Accountability rises. The premium shifts from doing the task to directing, auditing, + owning the outcome. Here's what surprised me most in spending time with super impressive and thoughtful professionals this week that are managing this transition for major enterprises: 𝘁𝗵𝗲 𝘀𝗵𝗶𝗳𝘁 𝗿𝗲𝗾𝘂𝗶𝗿𝗲𝗱 𝘁𝗼 𝗮𝗰𝗵𝗶𝗲𝘃𝗲 𝘁𝗵𝗲𝘀𝗲 𝗯𝘂𝘀𝗶𝗻𝗲𝘀𝘀 𝗼𝘂𝘁𝗰𝗼𝗺𝗲𝘀 𝗶𝘀𝗻'𝘁 𝗽𝗿𝗶𝗺𝗮𝗿𝗶𝗹𝘆 𝘁𝗲𝗰𝗵𝗻𝗶𝗰𝗮𝗹. 𝗜𝘁'𝘀 𝗵𝘂𝗺𝗮𝗻. 𝗔𝗻𝗱 𝗶𝘁 𝗿𝗲𝗾𝘂𝗶𝗿𝗲𝘀 𝗸𝗻𝗼𝘄𝗶𝗻𝗴 𝗵𝗼𝘄 𝘁𝗼 𝗺𝗮𝗻𝗮𝗴𝗲 𝗰𝗵𝗮𝗻𝗴𝗲 𝗰𝗼𝘂𝗿𝗮𝗴𝗲𝗼𝘂𝘀𝗹𝘆 𝗮𝗻𝗱 𝘄𝗲𝗹𝗹. Getting to real numbers in the form of more output, faster cycle times, and higher revenue per head requires courageous change management. 𝗧𝗵𝗲 𝘁𝗼𝗼𝗹𝘀 𝗰𝗿𝗼𝘀𝘀𝗲𝗱 𝘁𝗵𝗲 𝗲𝘅𝗲𝗰𝘂𝘁𝗶𝗼𝗻 𝘁𝗵𝗿𝗲𝘀𝗵𝗼𝗹𝗱. 𝗡𝗼𝘄 𝗰𝗼𝗺𝗲𝘀 𝘁𝗵𝗲 𝗵𝗮𝗿𝗱𝗲𝗿 𝘄𝗼𝗿𝗸: 𝗿𝗲𝗼𝗿𝗴𝗮𝗻𝗶𝘇𝗶𝗻𝗴 𝗮𝗿𝗼𝘂𝗻𝗱 𝘄𝗵𝗮𝘁 𝘁𝗵𝗮𝘁 𝗺𝗮𝗸𝗲𝘀 𝗽𝗼𝘀𝘀𝗶𝗯𝗹𝗲. 💭 Would love to know how people are thinking about the human/change management part of this puzzle.
2 Comments -
AFixt
Did you know? We created a repo that provides Gherkin-syntax versions of the ARIA Authoring Practices Guide (APG)? There are many uses for this. Two of our favorite reasons are: 1. To provide easy-to-follow descriptions for how common widgets should work 2. To provide a basis for QA automation of UI frameworks. https://lnkd.in/e6iMm4sv
-
CEQA.ai
Execution angle: CEQA/NEPA AI programs fail less when teams design from data relationships outward—not from prompts inward. If entity structure, governance ownership, and developer tooling are aligned first, AI outputs are more usable in real review workflows. Three current signals: 1) CEQ governance model for modern permitting technology: https://lnkd.in/g7JFvQk7 2) CEQ developer resources for NEPA/permitting systems: https://lnkd.in/g8jNZQ2J 3) CEQ data-standard entity integration model: https://lnkd.in/g6U8eaZN Practical takeaway for environmental teams: Define entity-level handoffs (project, process, document, decision) before scaling model usage. Structural clarity is what improves consistency, review speed, and defensibility together. #CEQA #NEPA #EnvironmentalCompliance #Permitting #AI #GovTech
-
arXiv Weekly AI Security Digest
📄 *Learning to Triage Taint Flows Reported by Dynamic Program Analysis in Node.js Packages* 👥 Authors: Ronghao Ni, Aidan Z. H. Yang, Min-Chien Hsu, Nuno Sabino, Limin Jia, Ruben Martins, Darion Cassel, Kevin Cheang 🧠 Why It Matters: This research explores the use of machine learning to prioritize potential vulnerabilities found in Node.js packages, making it easier for security analysts to focus on the issues that truly pose a threat. For business or enterprise security teams, this could mean reduced manual effort, faster identification of real threats, and ultimately, enhanced cybersecurity protection. Imagine a situation where your team doesn't have to waste time sifting through countless alerts about non-issues. Instead, they can focus on the actual vulnerabilities that need immediate attention. 🔗 https://lnkd.in/eTiqXVeA #CyberSecurity #arXiv #Research #ExecutiveTech #SecurityLeadership
-
BPB
Meet Sepehr Pakbaz, the author of 'Understanding Microservices with .NET', your end-to-end playbook for designing, building, and operating cloud-native systems that scale. 🚀 Sepehr Pakbaz is a cloud and AI solutions architect based in the San Francisco Bay Area with over 20 years of experience building large-scale distributed systems. Having worked for and with industry giants like Microsoft, Amazon, IBM, and Walmart, Sepehr has spent two decades guiding Fortune 500 companies through the complexities of modernization. He is a certified expert across Azure, AWS, and Kubernetes, and a dedicated community leader who bridges the gap between complex architectural theory and pragmatic engineering. A Look at the Chapters (Your .NET Microservices Roadmap) 👇 This systematic guide moves you from architectural theory to a fully functional e-commerce ecosystem: ➡️Chapters 1–4: Architecture & Design Patterns: Master the core principles of distributed systems and learn when to apply powerful patterns like Domain-Driven Design (DDD), Saga, CQRS, and Circuit Breakers. ➡️Chapters 5–10: Development & Implementation: Build a complete e-commerce application using .NET Web APIs and gRPC. Develop core services for Catalog, Basket, and Ordering while implementing asynchronous messaging with MassTransit and RabbitMQ. ➡️Chapters 11–13: Security, Containers, & Testing: Implement centralized security using OAuth2 and OIDC, containerize your services with Docker, and establish robust Test-Driven Development (TDD) strategies. ➡️Chapters 14–18: Deployment & Kubernetes Orchestration: Master CI/CD pipelines, Infrastructure as Code (IaC) with Terraform, and GitOps workflows. Learn to deploy, manage, and scale microservices on Kubernetes using Helm. ➡️Chapters 19–22: Operations & Future Trends: Achieve deep observability with Prometheus, Grafana, and OpenTelemetry, and explore the next frontier of Agentic AI microservices. This resource is the perfect playbook for .NET Developers, Solution Architects, and DevOps Engineers who are either starting their microservices journey or modernizing existing monoliths into cloud-native powerhouses. Want to have a sneak peek? Check out the free preview here! https://lnkd.in/gjyr9t38 Grab your copy of Understanding Microservices with .NET today! India: https://lnkd.in/g2b6PvwM Worldwide: https://lnkd.in/gwSGVEDS #DotNet #Microservices #Kubernetes #Docker #CloudNative #DevOps #SoftwareArchitecture #CSharp #NewBookAlert #BPBOnline
10 Comments -
Trailhead Technology Partners
Microsoft's new Microsoft Agent Framework (Preview) for .NET shows how developers can orchestrate LLMs, tools, memory, and planning/execution components to build multimodal, tool-enabled AI agents that integrate with model providers and external services. #DotNet #AI #GenerativeAI https://hubs.ly/Q03LJHrv0
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content