Danilo Poccia
banner
danilop.bsky.social
Danilo Poccia
@danilop.bsky.social
1.4K followers 830 following 480 posts
Chief Evangelist (EMEA) @ AWSCloud. Serverless, IoT, AI/ML. Few pics, some music. My opinions. Complexity is a science. He/him.
Posts Media Videos Starter Packs
Reposted by Danilo Poccia
Our latest videos are out!
1️⃣ From Prototype to Production with Strands Agents & Amazon Bedrock AgentCore - @danilop.bsky.social
2️⃣ My AI Coding Partner: Building the London Improv Calendar on AWS - Diana Ionita
youtube.com/c/AWSUserGro...
#AWS #AI #AmazonBedrock #Agents #Serverless #AWSCommunity
Reposted by Danilo Poccia
I just mentioned an AWS tool in my talk and the next speaker Ricardo Sueiras, is going to demo it. He's doing a great job of explaining how to set rules and context when generating code.

"The more effort you put in with the prompt the better the output."
Learning about the latest technologies at @codemotion.bsky.social 😂 Seriously, lots of nostalgia and curiosity from the audience… depending on their age 😅 #Codemotion #Milan
Reposted by Danilo Poccia
New milestones! The AWS News Feed now has over 2k digest subscribers and sent out over 20k emails in the past 30 days ✉️ 💌 📈

The AWS News digests are the best way to stay on top of everything new in AWS, especially with Re:Invent 2025 approaching. Not subscribed yet? Do it here: buff.ly/Qe3cvlJ!
On my way to Milan for Codemotion, see you there tomorrow!

#CodemotionMilan
Excited to be back at Codemotion! My session will be on Tuesday morning (just after the not-to-be-missed keynote by Enrica Porcari, Head of IT, CERN).

From Prototype to Production: Context Engineering and Memory Patterns for AI Agents

#CodemotionMilan
It was so much fun to be there! I met friends I haven’t seen for years. Thanks for having me!
Had a great time at our first AWS AI in Practice meetup last night. Thanks to all! Big thank you to @danilop.bsky.social & Diana Ionita for 2 great talks! Cannot wait for the next one - 12th Nov. www.meetup.com/awsuguk-ai-i...

#AWS #AI #AmazonBedrock #Agents #Serverless #AWSCommunity #APIGateway
Deploy production AI agents on Amazon Bedrock AgentCore in minutes!

This comprehensive guide by Mike Chambers shows how to transform your laptop prototype into enterprise-grade AI with persistent memory, secure code execution, and automatic scaling.

buff.ly/TNZpXBu

#AWS #AI #GenAI #AIagents
AWS Builder Center
Connect with builders who understand your journey. Share solutions, influence AWS product development, and access useful content that accelerates your growth. Your community starts here.
buff.ly
Reposted by Danilo Poccia
Nice example of using Strands Agents with LiteLLM! Thanks for sharing 👏
Nice example of using Strands Agents with LiteLLM! Thanks for sharing 👏
Wow nice, thanks for sharing!
Reposted by Danilo Poccia
The stable release of Python 3.14 is out now! Go, go, go update! 🙌

discuss.python.org/t/python-3-1...
Reposted by Danilo Poccia
Next week - Wed 8 Oct London.
🎙️ @danilop.bsky.social - From Prototype to Production with #StrandsAgents & #AmazonBedrock #AgentCore.
🎙️Diana Ionita - My #AI Coding Partner: Building the London Improv Calendar on #AWS
www.meetup.com/awsuguk-ai-i...
#AWSCommunity #Serverless #DynamoDB #APIGateway
All patterns shown are framework-agnostic and apply to any LLM backend. The open source Strands Agents framework and Amazon Bedrock AgentCore are used for the specific implementation.
Through hands-on demos, we'll demonstrate the implementation and production deployment of context management patterns such as summarization and memory hierarchies separating working memory from episodic/semantic stores.
...
While prototypes can naively append all conversation history and tool outputs, production agents face context window limits, degraded performance from "context rot", and growing costs as conversations lengthen.
...
Context engineering, deciding what goes into the model's context window at each step, determines whether your agent fails after 10 interactions or completes 1000-step tasks reliably.
...
AI agents are a new class of software applications that use AI models to reason, plan, act, learn, and adapt in pursuit of user-defined goals with limited human oversight.
...