AWS Developers’ cover photo
AWS Developers

AWS Developers

Technology, Information and Internet

Connecting developers to the resources, community, and technical guidance that turn ideas into reality with AWS.

About us

Website
https://aws.amazon.com/developer/
Industry
Technology, Information and Internet

Updates

  • You can now run AI models as Lambda functions! ⚡ Until now, putting a model on a GPU meant running GPU servers, picking instance types, installing drivers, scaling up and down and keeping them busy enough to justify cost. With Lambda Managed Instances, you deploy the model as a function instead. Just tell the function how much GPU memory it needs and Lambda picks the instance, launches it and scales within the limits you set. GPUs are built to do thousands of calculations at once, so they perform best with many requests together. Each execution environment takes multiple concurrent requests, and a model server like vLLM batches them on the GPU. One GPU, several times the work. So what can you run? → Strands Decider for fast classification and routing → NVIDIA Nemotron Parse for documents and structured data → NVIDIA Nemotron Embed for search Each one is its own function on its own GPU.

  • AWS Developers reposted this

    View profile for Matt Garman
    Matt Garman Matt Garman is an Influencer

    From AWS Heroes who've been building on AWS for years to university students heading to their first re:Invent, our builder community will be together in Las Vegas in less than two weeks. Our AWS Heroes and AWS Community Builders are leading sessions all week on what they've learned. And for the first time, 60 students on our All Builders Welcome grant, chosen from thousands of applicants, will be there to learn from them.  Hope to see many of you there. If you can't make it, join one of 46 Watch & Build Parties around the world or a watch party at our Builder Lofts in San Francisco, Berlin, and São Paulo. https://lnkd.in/ePJcT5rQ

    • No alternative text description for this image
  • Starting a new conversation with an agent shouldn't mean starting from scratch. MongoDB and Strands Agents are introducing 𝚜𝚝𝚛𝚊𝚗𝚍𝚜-𝚖𝚘𝚗𝚐𝚘𝚍𝚋, a Python integration that lets your agents remember useful facts from one conversation to the next. Your agent gets two kinds of memory: 1. Session memory, which saves the conversation you're in the middle of so a restart or a move to another instance doesn't lose your place. 2. Long-term memory, which keeps what's worth holding onto after that conversation ends. You can tell your agent your project's integration tests use a separate database and later, in a new conversation, ask what test setup the project needs, and it can bring up that database without you repeating it. When your agent needs a memory, it searches in plain language. MongoDB Atlas handles the embeddings and uses Vector Search to find memories that match in meaning, even when the wording is different.

  • Not heading to Las Vegas for AWS re:Invent this year? You can still be part of it. 🌍 During re:Invent week, November 30 to December 4, AWS community leaders around the world are hosting re:Invent Watch & Build parties in their cities. These free, local events are run by AWS User Group Leaders, Student Builder Group Leaders, Community Builders, and Heroes. At a Watch & Build party, you can: - Watch selected re:Invent content together - Build hands-on with AWS services - Meet builders in your community Parties are happening from Colombia to Vietnam, the United Kingdom to Kenya. Find one near you and register on Builder Center. Learn more with the link in the comments 👇

    • No alternative text description for this image
  • View profile for Brooke Jamieson
    Brooke Jamieson Brooke Jamieson is an Influencer

    Byte-sized tech tips for AI + AWS

    big changes for the Well-Architected Framework today! It's our first big change in 2 years, and here's why it's important for developers. Every best practice is now something you can check in your config or telemetry - and that means an AI agent can check it too! Account-wide questsions lke root account protection, sign-in, and logging now live in a brand new section called 'Cloud Foundations'. This means your org can just answer them once, so your app's review only asks about your app. There's also new best practices too, like backups a breached admin can't delete. Here's a link to the blog: https://lnkd.in/gUZcEWbd When's the last time your team did a Well-Architected review? 🐈 This is a clear and conspicuous disclaimer that I am an AWS employee and all opinions are my own 🐈

    • Cheat sheet about the AWS Well-Architected Framework update, with Brooke Jamieson standing in the middle. Title: AWS rewrote Well-Architected so an agent can check your architecture. A tag reads: Launched 16 Nov 2026. On the left, two stacked boxes, with an arrow pointing up from the bottom box to the top one. Top box: Reviewing your app? Questions now cover only the app itself. Bottom box, highlighted as new: Asked once per org: root account, sign-in, central logging, networking, billing. Often answered by your platform team. The arrow says: every app inherits these. On the right, a box titled Which advice stayed? asks: Can config or telemetry show it? It splits two ways. Yes: kept. No: retired: strategy, people and process. Below it, a box titled New best practices lists four, each with a tick: secure coding rules, checked in code review; backups a breached admin can't delete; AI helps analyze cost, people make the call; measure resource use per active customer.
  • How much AWS access does your AI coding agent really need? Usually just enough for the one job you gave it. That's least privilege, and it matters more once an agent is running commands for you. Your agent will try to follow what you ask in the prompt, but AWS only goes by the agent's permissions. Scope its IAM role to the actions and resources the task needs. You can get started by naming the tasks it needs to do, then matching the access to them: → Observe, like finding out why a function failed. Read its settings and logs. → Build, like updating a function. Change only that function and its deployment files. → Destroy, like removing an old experiment. Run a separately reviewed cleanup, limited to that experiment. → Changing access, like creating roles or editing policies. Review these closely, because an agent that deploys apps shouldn't be able to change its own access. For production, let the agent investigate, and send changes through a reviewed deployment pipeline with its own role. The slides cover five steps, from checking your agent's identity to fixing AccessDenied errors. The full guide is linked below👇

  • View organization page for AWS Developers

    74,959 followers

    If you're a student who's ever wanted to get people on campus building together, an AWS Student Builder Group is a good place to start. Applications are open now. Over 12 months, you'll host events, ship projects, share what you learn, and grow a community of student builders at your school. Along the way, you'll earn AWS credits and certification vouchers to keep building, plus a QA learning license, badges, and swag. You won't be doing it alone, either. You'll have a dedicated Account Manager, monthly check-ins with the AWS team, and a Slack community of group leaders in 68+ countries. There's one application with two close dates, depending on how you want to get involved: → Join a group that's already open by Sunday, Nov 15, 2026, at 11:59 PM PT → Start a new group for the next cohort by Sunday, Jan 24, 2027, at 11:59 PM PT. To apply, you need to be 18 or older, enrolled with at least 12 months left at your institution, and ready to run at least 6 events over the year. You don't need any AWS experience, just the drive to build and bring people along with you. Link in the comments 👇

Affiliated pages

Similar pages