Tobias Schmidt
LightNews LightNews
banner
tpschmidt.com
Tobias Schmidt
@tpschmidt.com
680 followers 64 following 1.1K posts
Helping aspiring engineers master the cloud 👨‍💻 Freelance Software Engineer ✍️ Book #1: http://awsfundamentals.com 📕 Book #2: http://cloudwatchbook.com Learn AWS for Free: https://awsfundamentals.com/newsletter
awsfundamentals.com
Posts Media Videos Starter Packs
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 2d
But the environment is 100% predictable.

Just learned about this through a case study from Piotr Migdał, founding engineer at Quesma:
awsfundamentals.com/blog/sandbo...
Running Untrusted Code Safely at Scale with AWS Lambda
A case study on how Quesma built a secure, isolated execution environment on AWS Lambda to safely run untrusted, AI-generated code in production.
awsfundamentals.com
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 2d
Example: execute generated code inside Lambda, locking it down in a VPC with no public internet. If you need S3 access, it can happen through a VPC endpoint. Lambda’s hard timeouts keep "rogue" code from doing bad things. Dependencies still need to be pinned, sure.
1
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 2d
Lambda has tons of use cases, but one I've missed: using it as some kind of sandbox for running AI-generated code.
Lambda's isolation and scaling are a solid fit for this problem.
1
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 3d
Btw I'm Tobi, a full-stack engineer who's broken (and fixed) plenty of AWS stacks in production. If you want a clearer picture of how AWS services fit together, I put together some free animations that break it all down.
awsfundamentals.com/animations
Animations | AWS Fundamentals
Learn visually with our animated guides to AWS services.
awsfundamentals.com
1
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 3d
Here's a small article about things I've learned and how to explore this by yourself, as code is included and ready to deploy to your own account ✍️
awsfundamentals.com/blog/cloudw...
How to Use AWS CloudWatch Application Signals with OpenTelemetry on ECS Fargate and Lambda
This guide shows how to connect CloudWatch Application Signals with OpenTelemetry. See simple steps for ECS Fargate and Lambda. Example code included. Get clear metrics and traces fast.
awsfundamentals.com
1 1
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 3d
Still has some quirks and found a few things that are not quite right in the documentation and/or are not easy to configure, but I like where AWS is heading with CloudWatch! 💛
1
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 3d
• Custom business logic spans for detailed application code visibility 🛠️
• Real-time service maps in CloudWatch Application Signals 🗺️
• Transaction Search to follow individual requests across your stack 🔎
1
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 3d
• Automatic service discovery that maps your application topology in real-time 📡
• Request tracing through ECS containers to Lambda functions 🕵️
1
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 3d
Then you can auto-instrument your app with one line of code and everything is ready to use.

What you'll get, mostly out of the box:
1
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 3d
For Fargate, you basically just need to run the CloudWatch agent as a sidecar. The image is built by AWS and comes ready to use. Just need to inject a few environment variables and you're ready to go.

For Lambda, you can use a Lambda layer that is also developed by AWS.
1
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 3d
Setting up an OpenTelemetry & CloudWatch-based observability solution with Fargate and Lambda is quite simple.
It also automatically integrates with Application Signals!

Played around with it a lot in the past few days and I like it.
1 1 2
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 5d
🎁 Bonus: The service configuration can also take care of setting up auto-scaling or using the Spot capacity provider.

Bootstrapping Fargate tasks was never that simple nor cheap! 🤩
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 5d
This means you'll just point to your repository's Docker file(s) and SST will build & upload images on demand.

No need for custom scripts.
1
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 5d
• a VPC with private & public subnets
• an ECS service, with a corresponding cluster & a task definition
• the task & task execution roles for ECS
• an API GW that routes requests through CloudMap & a VPC integration

SST will also handle the Docker magic.
1
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 5d
Setting up an exposed Fargate task with ECS is still not a simple task. With sst.dev, this is enough to bootstrap and wire everything 🪄

What's behind these few lines?
1
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 6d
Afterward, you can try out Teleport for free for 14 days to get your own perspective and experience on their service:
fandf.co/4nJSbwR
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 6d
If you're just starting out or if you want to take your agent setup to a professional level in your company, have a read of Teleport's in-depth guide on MCP security. You'll definitely learn a lot:
fandf.co/3IFbAQF
1
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 6d
In a nutshell 🥜
Each LLM or service gets just the access it needs, only when it needs it. Fresh, short-lived credentials. All actions logged.
1
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 6d
Good news: you don't need to build this yourself.
Teleport.sh solves all of this for you!

Big thanks to them for partnering on this post and introducing me their platform! 🤝
1
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 6d
So what do we need to make this better?

1. Dedicated identities for the LLMs/MCP servers 👥
2. Tailored permissions, only for actions that are necessary 📝
3. Short-lived credentials 🔑
4. Auditing of actions taken ✏️
1
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 6d
When taking it further, thinking about multi-agent scenarios where agents are fanning out calls to other agents, this traceability issue becomes even worse.
1
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 6d
The problem: if you run them locally via your own assumed role, or even remotely in some instance or Lambda function, you always have the issue of sticky long-term credentials.

What's even worse is that there's zero traceability out of the box.
1
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 6d
Most people I know worry about their AWS accounts' security.
With no spending cap, this is a valid concern.
Today, with LLMs & MCP servers, security becomes even harder.

I love using MCP servers for AWS tasks, especially with the large set of official ones provided by AWS.
1
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 6d
Worth to check out the docs from time to time as they are crystal clear.
docs.aws.amazon.com/AmazonCloud...
tpschmidt.com
Tobias Schmidt @tpschmidt.com · 6d
• ⚡️ 𝗟𝗮𝗺𝗯𝗱𝗮: just pull in the ADOT layer, set a few environment variables and you're good to go.
• 🤖 𝗘𝗖𝗦: either bundle the auto-instrumentation dependency into your app or use the pre-built docker layer as an init sidecar.

Yes, it's that simple.
1 1