Raghvender Arni’s Post

TL;DR: Step functions continue to be an excellent foundation for GenAI Apps built on Amazon Web Services (AWS) Bedrock In one of my previous posts I highlighted how Step functions are used in almost all GenAI apps my team built: https://lnkd.in/e3HDTzMX As a recap on why?: It has become common practice to break a broader GenAI problem apart into multiple steps, like decomposing the problem, using different models to solve the each of the problem parts, handling edge cases, generating, validation and merging responses, etc. What better way to do all that than use a robust workflow engine. Yes, some of these will be solved via agentic workflows which will be magical but will also not be predictable. https://lnkd.in/esshzKea For those times you want predicable workflows, Step Functions are the best. Why 1/Visual Workflow Design 2/Serverless – Scale up and down on a dime 3/Reliability and Fault Tolerance 4/Integrated Security 5/Logging and Tracing including X-ray support 6/Local Testing 7/East connectivity to key AWS services and external ones using AppFlow It was awesome to see Clare Liguori put together this tech guide that shows some great patterns for Prompt Chaining workflows for Bedrock. https://lnkd.in/efryACXD -- Model invocation -- Prompt templating -- Sequential chains -- Parallel chains -- Conditions -- Maps -- Chain prompts and other AWS services -- Validate output and re-prompt -- Wait for human input If you are not using Step Functions for your GenAI Apps, you should!

Love this example of how Bedrock plus Step Functions helps you balance the power of GenAI with the determinism and control given you by step by step workflows.

The latest update for #Logzio includes "Logz.io Upgrades App 360, #Kubernetes 360 with AI Assistant, New Tracing Quickview" and "#DevOps and #SRE Metrics: R.E.D., U.S.E., and the 'Four Golden Signals'". #Logging https://lnkd.in/dGcUEWi

🌟 Missed .NET Conf? Catch up on my session now! 🌟 Microservices are transforming modern app development, and with the power of AI and containerization, we can take them to the next level! In my session, I explored: 🔹 Building microservices with .NET 9 – Prioritizing modularity, scalability, and maintainability. 🔹 Integrating AI using Azure OpenAI – Enabling intelligent decision-making in microservices. 🔹 Deploying AI-powered microservices with Docker – Simplifying containerization and ensuring cloud readiness. Don't miss the opportunity to learn how to create smarter, scalable applications. Watch the full session here: https://lnkd.in/eXiMSXJM Let’s bring intelligence to microservices, one container at a time! 🚀 #DotNetConf #Microservices #AI #AzureOpenAI #NET9 #Docker #Containers #CloudComputing #DeveloperCommunity #KeepLearning #KnowledgeSharing #TechTalk #LearnAndGrow #CodeWithAI #AIForDevelopers #SharingIsCaring

Amazon Web Services Wants to Host Custom Models Amazon Web Services (AWS) announced the preview of Custom Model Import for Amazon Bedrock. Amazon Bedrock allows users to import customized models for generative AI application development, leveraging their prior model customization investments. This feature supports architectures like Llama, Mistral, and Flan T5, allowing users to import and access customized models on demand. Amazon Bedrock has launched a new feature allowing customers to import customized models outside the platform. This enhancement of the user experience for application developers will allow them to access these models on-demand through Bedrock's invoke model API. The feature is currently in preview in the US-East (North Virginia). AWS wants to be the infrastructure that powers generative artificial intelligence, whether it is custom models or models aggregated from various startups and incumbents. AWS offers developers API access to AI models that can speed up the delivery of apps or software built for teams and/or businesses. Adding API is another potential time taker for software engineers, but hosting custom models via Amazon Bedrock. What other cloud software enables startups or incumbents to access their own models or models from other companies conveniently? #ai #models #aws #custom https://lnkd.in/gJdTjka5

𝗪𝗵𝗮𝘁 𝗿𝗲𝗮𝗹 𝘃𝗮𝗹𝘂𝗲 𝗰𝗮𝗻 𝘁𝗵𝗲 𝗻𝗲𝘄 𝗜𝗻𝗹𝗶𝗻𝗲𝗔𝗴𝗲𝗻𝘁 𝗳𝗲𝗮𝘁𝘂𝗿𝗲 𝗯𝗿𝗶𝗻𝗴 𝘁𝗼 𝗮𝗴𝗲𝗻𝘁 𝗱𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁? If you've built an agent in Amazon Bedrock, you know the drill: You need to create an agent role, configure the agent—set it up, define instructions, configure action group tools, grab the agent ID, assign an alias, and finally obtain the endpoint to integrate it into your use case. This process often involves tools like 𝗖𝗹𝗼𝘂𝗱𝗙𝗼𝗿𝗺𝗮𝘁𝗶𝗼𝗻, 𝗔𝗪𝗦 𝗖𝗗𝗞, or 𝗯𝗼𝘁𝗼𝟯. But now, with 𝗜𝗻𝗹𝗶𝗻𝗲𝗔𝗴𝗲𝗻𝘁, you can streamline all of this 𝗼𝗻-𝘁𝗵𝗲-𝗳𝗹𝘆 𝘄𝗶𝘁𝗵 𝗮 𝘀𝗶𝗻𝗴𝗹𝗲 𝗔𝗣𝗜 𝗰𝗮𝗹𝗹!! 🎯 You can dynamically configure your agent at runtime, specifying foundation models, instructions, action groups, guardrails, and knowledge bases without needing pre-configuration. This means you can rapidly build, test, and experiment with agents in real-time, enabling faster iterations and innovation. 𝗦𝗵𝗼𝘂𝗹𝗱 𝗶𝘁 𝗯𝗲 𝘂𝘀𝗲𝗱 𝗶𝗻 𝗽𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝗼𝗻? I’d say 𝗡𝗢 🚧 While InlineAgent is great for rapid experimentation and prototyping, agents intended for production should still be created using Infrastructure as Code (IaC) tools. Check out the link below to know more: #agenticworkflow #agentai #generativeai #amazonbedrock #amazonsagemaker #future https://lnkd.in/ePSc2FfW

𝐇𝐨𝐰 𝐃𝐨 𝐘𝐨𝐮 𝐓𝐫𝐚𝐜𝐤 𝐚𝐧𝐝 𝐌𝐨𝐧𝐢𝐭𝐨𝐫 𝐘𝐨𝐮𝐫 𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐯𝐞 𝐀𝐈 𝐀𝐩𝐩𝐥𝐢𝐜𝐚𝐭𝐢𝐨𝐧𝐬 𝐢𝐧 𝐏𝐫𝐨𝐝𝐮𝐜𝐭𝐢𝐨𝐧? Watch the demo video to see how Langfuse optimizes the entire app’s lifecycle.....from development to monitoring and testing! Langfuse (YC W23) offers comprehensive LLM observability for generative AI applications by allowing you to instrument your app and ingest traces for real-time monitoring. Its UI enables inspection and debugging of complex logs, while prompt management features allow seamless versioning and deployment of prompts. With Langfuse, you can test and iterate prompts in the LLM Playground, track critical metrics like cost, latency, and quality through dashboards, and evaluate LLM completions with model-based evaluations and user feedback. Langfuse also supports A/B testing and benchmarking performance with datasets before deployment. You can get started quickly with Langfuse Cloud, which offers a generous free-tier, or opt for self-hosting via Docker on platforms like AWS, GCP, or Kubernetes, making it easy to manage gen ai applications. Ps : I am bad at editing videos ....please don't mind lol #langfuse #ai #production #llmops

Amazon Q for Business is your fast-ramp to offer generative AI driven services using your business data to your customers. This solution is a template for building a custom UI for using Q for Business on your enterprise data sources, with security and guardrails under your control. With this solution you can offer an interface quickly while using your organization's authentication and authorization mechanisms. #awsishow #amazonq

What's the outcome of GenAI BuildCamp 2024? AI BuildCamp will equip you with skills in building full-stack Retrieval Augmented Application (RAG) including SSO authentication with Cursor IDE, build and deploy to Google Firebase. You will be working on real-world use case of financial loan documents which are typically 50 - 100 pages long. This is the depicted architecture for the app. After the program, you will be experienced with building RAG and full-stack web application using an AI-assisted IDE like Cursor. You can use the skill you learn from the course for using other AI models and other cloud platforms such as Azure, AWS. Bonus - you will be invited to join a live session with Vatsal Sanghvi (https://lnkd.in/gQA3_dmP) - serial builder and founder of 1811Labs from Bangalore - to learn the playbook with customer acquisition on Product Hunt & Reddit. This will be valuable for those of you who want to launch what you build through out the program and make money from that. We are closing registration in 3 days. Only 4 spots left. Get your spots today https://lu.ma/sjae5p4o

Kickstart your AWS journey with these hands-on projects! 1. Simple Static Website: Begin with the basics by hosting a static website on Amazon S3. For instance, create a personal portfolio or a blog site using HTML, CSS, and JavaScript. 2. To-Do List Web App: Build a to-do list web app using AWS Amplify. This serverless approach will give you a taste of front-end and back-end development without managing servers. 3. Image Recognition: Dive into AI using Amazon Rekognition to create an image classification project. Take a photo and let your application identify objects, people, text, scenes, and activities. 4. Sales Data Dashboard: Work with data using Amazon QuickSight to build a sales data dashboard. Experiment with different visualization techniques to present your data effectively. 5. Customer Service Chatbot: Experiment with conversational AI. Use Amazon Lex to build a simple customer service chatbot that can answer common questions about your business. Each of these projects can be a stepping stone to bigger and better innovations. Start exploring!