DEV Community

Cover image for My AWS Summit Toronto experience and GenAI

My AWS Summit Toronto experience and GenAI

I have been experimenting and working with AWS for many years now. Way back when, I started out by purchasing a training course put out by Yan Cui (featuring his cat) and have learned so much since then. I was fascinated with the idea of the scale and flexibility possible using cloud services and the things you could do with serverless services like AWS Lambda and S3.

As my journey has progressed over the years I have interacted with so many amazing people and learned so much from them. I have joined a number of communities and regularly participate in online sessions, webinars, and other events to learn whatever I can. I have spent the most time in the last couple of years hanging out at the Believe In Serverless Community. It is a great place to discuss and share ideas on serverless approaches in the cloud and there are tons of great people hanging out there.

One thing I haven’t been able to do though is meet most of the people in the AWS (and related) communities in person. This past week I was able to attend the AWS Summit in Toronto, Canada and had a great experience meeting so many people and learning new things. The experience has really energized me to do more in the future. I have attended similar conferences in the past like Java One but am very glad to have had this experience.

The AWS Summit Toronto

Toronto is about 4 hours drive from Ottawa where I live but it’s the closest AWS Summit location to me. I will be attending AWS Re:Invent for the first time this year but when I saw the opportunity to take part in something earlier I decided it was a no brainer. I’m glad I did as I met many local people from the AWS Community and many of them may not be able to be in Las Vegas in December as well.

I have been super busy with a project at work the last couple of weeks and it was a great time to be able to head out and spend some time not thinking about my work projects. I was able to stay in Toronto the day before and after the summit and took these days to relax, do some workouts, and start to feel normal again.

I ended up picking up my badge for the summit the day before to save time on summit day as I had a full plan of talks to attend. I have never been in the Metro Toronto Convention Centre before and kind of got lost finding my way to the badge pickup location. I did eventually find it and it all felt a lot more real seeing the huge setup that was there for the AWS Partners Summit. I didn’t attend that this year but maybe next time as our company is part of the program.

AWS Summit Toronto Badge

I didn’t sleep super well on the night before the summit as I didn’t want to be late and miss anything. I had a much better journey on summit day and used the pathways above the rail tracks in the downtown area to go to directly to the Convention Centre. I was inside and waiting about 45 minutes before the first session I wanted to attend.

As I was researching the session catalog in the days before the summit it was clear there would be a real focus on GenAI topics. This is really no surprise as we’ve seen this from AWS (and the tech world in general) for a long time now. Many people don’t like this but it’s the reality of where we are at today and the sheer amount of interest (and promise) in this space. I was able to pick out 7 sessions I wanted to attend that didn’t overlap. Three of these were GenAI related, one was the keynote, one was related to my favourite subject of serverless computing on AWS and there were two others. I think I had a good mix of what interests me today.

It’s all about community and personal interactions

The real highlights for me from the AWS Summit Toronto were meeting and seeing presentations from AWS people who I have learned so much from over the years and meeting so many AWS community members.

Tracy Wang from AWS organized a mixer at the end of the summit day at a nearby hotel where I got to finally meet so many people from the local community. Many of these people live rather close to me but I have not met them in person. It’s just not the same meeting someone on Zoom or other online video platforms as seeing them in person. Meeting AWS Heroes like Cyril Bandolo and Andrew Brown was a real honor. I have spoken with them online before but in person is so much better. I also met so many other really interesting people that seem very passionate about AWS. I will probably miss mentioning some of them but meeting people like Cyril, Andrew, Rola Dali, Bansi Delwadia, Hubert Chan, Farah Abdirahman, Tracy Wang, Aaron Brighton, Varun Kumar Manik and more was a real honour and inspires me.

For my work with Ciena, we have many colleagues in locations around the world including in many cities in the US as well as our offices in Gurugram (Delhi), India. I have had the pleasure to travel to India a number of times to see our offices there and meet almost all the people I have worked with. These experiences have greatly enhanced my interactions with all these people on all the other times when we’re on opposite sides of the world working together on projects. Once you see people in person and see their mannerisms and how they respond in person it is so much easier.

Even in cases with people who I work with based at our offices in Ottawa, Canada so much has changed in the last number of years. I have been at this for many years and in the earlier part of my career I learned so much from the more experienced people in my company by just hanging out at the office and in the coffee rooms. I would not be the architect, developer and overall person I am now without these experiences. It has to be difficult for recent people entering our field. Real human interactions like meeting people in person is really needed to build good relationships IMO. People from the community coming together at events like the Toronto Summit is very important and we really need to try and do more of this.

The Community Mixer

Highlights from the sessions I attended

AWS is really doing an all-out effort to gain traction in the industry in the Generative AI space. Just a few years ago I was working on a more traditional AI/ML team in an MLOps role using AWS and Kubernetes. Back then Natural Language Processing (NLPs) were just one subfield we were working on that didn’t seem to have much more promise than others. Of course that all changed thanks to the development of transformers and the ability to train huge language models. The rest is history and the world has been transformed with GenAI.

DAT302 (Building cost-effective RAG-based apps with Amazon Aurora)

Many companies are starting out with GenAI by using Retrieval Augmented Generation (RAG) approaches where their specific data. Using RAG with generic Large Language Models makes it possible for companies to get real wins with GenAI approaches for minimal costs and effort. Using RAG along with Embedding Models, Vector databases and more can still really add up in costs if not optimized. The session DAT302 (Building cost-effective RAG-based apps with Amazon Aurora) presented by Shayon Sanyal and Aditya Samant included a number of good tips on optimizing spend with these approaches.

Some ideas included using multiple levels of models to reduce costs like first passing raw data to a smaller and cheaper model to summarize and then only passing this smaller set to the more expensive and powerful models. There were also specific settings with Aurora databases that work best with RAG.

SVS201-R (Building Production-ready Agentic AI Architectures with AWS Serverless)

This is honestly one of the two sessions I was most looking forward to. Anton Aleksandrov and Heeki Park are excellent presenters and I have watched and read so much of their content over the years. We had a great discussion about the approaches you can use for building Agents using tools like the Amazon Bedrock, the Strands SDK, AWS Lambda, the Elastic Container Service (ECS), Step Functions, and of course the new Bedrock AgentCore suite of tools.

Anton and Heeki discussed how agents communicate with each other, require memory to work together in an Agentic fashion and how Agents are not really all that different than micro-services. The audience got lots of high level ideas on how best to approach agents and all the tooling AWS offers to help. Seeing these guys in person and chatting with Anton were highlights for me.

Serverless Agents with Anton and Heeki

The Keynote with Andrew Kent Warfield

I know some people find keynotes at conferences like this boring but I was glad to able to be in the room for this one. It was great to hear how someone who grew up in Canada has become a key part of AWS. As with most of the rest of the conference there was a heavy focus in the Keynote on the various GenAI products AWS has been working on but there were also some nice stories about the role AWS has played with many Canadian companies and driving innovation here.

SEC302 (Building a Secure EKS Framework For Financial Services)

I have been setting up and administrating Kubernetes clusters and using the Elastic Kubernetes Service (EKS) on AWS for a number of years now and we run multiple clusters in production. I wanted to go to this session to understand what the recommended approaches are to make sure everything is secure. Raj Bagwe and Preetam Rebello did a great job walking through many best practices to ensure security in your installs.

Items like using private connectivity, enabling control plane logs, using encryption for everything possible, setting up clusters using the least privileges possible, using EKS Pod Identities, Network Policies, OPA, and other tools, looking at using Service Meshes like Istio, and so many other great tips were presented. In my case it was great to see that we are already in really good shape as we pretty much used all the approaches discussed. Once the slides are available I will go over everything in detail to make sure and I did provide lots of comments and feedback during this session.

EKS Security Tips

ANT401 (Architectural Patterns for Near Real-time analytics on AWS)

In the GenAI and telecom equipment space where my work is focused we have a number of use cases for ingesting large amounts of network data and the using it to build applications. I have worked on multiple project with this kind of approach and have used Kinesis streams as well as custom solutions.

Mazrim Mehrtens and Vinodh Sadayamuthu discussed some best practices they recommend to AWS customers including the use of Kinesis or Amazon Managed Streaming for Apache Kafka (MSK) for ingestion and then tools like the Managed Service for Apache Flink for super performant data processing. We have looked at using Flink in the past but based on this presentation I plan to take another look. There were some great tips and I am really looking forward to looking at the slide and the example github repos that were discussed.

In the GenAI space one real use case I see here is building models that can predict or detect anomalies by doing inferences with real time data on GenAI models and also the constant retraining and updating of these models. Being able to use some of the techniques discussed by Mazrim and Vinodh could really help.

Streaming tips

AIM301 (Deploy, secure, and observe Production-grade AI Agents with AgentCore)

This is the other of the two sessions I was most looking forward to. The announcement of the preview for AgentCore at the AWS Summit New York was a super interesting development for me. There are so many tools and libraries out there for working with GenAI but there really are a smaller number that seem ready to be able to help deploy real Agentic AI apps at scale with proper security and reliability.

Fei Yuan and Satveer Khurpa did a great job describing some key components of AgentCore in detail. Fei especially did a great job white-boarding some of the key components and flow of the AgentCore Identity piece. In order for Agentic AI apps to be production ready I believe security has to be front and centre. I did find the details of how the Identity module works to be rather complicated but got a really good idea of how it works from this session. I also found out they have an AgentCore Discord server so I joined that after the session and am really looking forward to interacting with Fei and other people working on AgentCore.

On the way back from Toronto on the train I was watching more videos on AgentCore and am planning on my next project migrating some existing code to run with AgentCore. I really think this set of tools (along with the Strands SDK) will play a big part for the future of companies deploying Agentic Apps in production.

Agentcore in production

SVS203 (Unlock Productivity: Tools and Techniques for Serverless Developers)

At heart I am a serverless-first person and you can't think about serverless on AWS without people like Julian Wood. I was not able to attend the full session as it overlapped with the AgentCore one but I was lucky enough to hear Julian Wood go over some great improvements in the Serverless space on AWS. Chris McPeek and Janak Agarwal also were part of this session and demonstrated many of the other super useful features AWS has been putting out in the serverless space.

At the end of the session, Julian showed a great demo of using Q to build a serverless app for him including SAM templates, running locally via SAM and also building a front-end for an app in seconds. It is truly amazing (and somewhat scary) what GenAI tools like Q can do and they are only going to get better.

A highlight for me was getting to chat with Julian for a few minutes after the session. I have attended so many streams where he presents all the latest in serverless and has all the key people from AWS with him.

Julian Wood on Serverless

Tips for others on attending summits and conferences

Some tips for people attending this summit (which likely applies to others as well)

  • If you’re travelling, book your (refundable) hotel well in advance. I’m kicking myself as I had a number of decent priced options back in June when I started planning but didn’t book and got stuck with a rather expensive (and non-refundable) booking maybe due to the Toronto International Film Festival being at the same time.
  • Scout out the summit location ahead of time if possible. I’m glad I did this the day before on badge pickup day and knew much better where to go on summit day.
  • Bring your own refreshments if you can. There were some provided as part of the summit but others were available at a rather high cost.
  • Plan out which sessions you want to see and make sure to show up at the room 30 minutes before as all the ones I attended were full by the time the door opened to let us in.
  • Ask questions during the sessions, share your insights, and introduce yourself to the presenters after if there is time and opportunity. I really got a lot out of personally saying hi to so many people that I have attended online sessions before and read content from.
  • Attend the keynote! It may sound boring but keynotes at conferences like this are always great to see and you get a much better sense of the overall purpose and direction of the company and domain.

Conclusion

When I first thought about attending the AWS Summit Toronto I wasn't really sure if it was would be worth it. Since the summit itself was free I wasn't expecting too much compared to something like Re:Invent but I was wrong on that front. I gained so much in knowledge and interactions on this trip and hope to do this again in the future. Now I'm looking even more forward to Re:Invent where it will be all of what I experienced but times 3 or 4. I'm also invigorated to do more and learn more.

For more articles from me please visit my blog at Darryl's World of Cloud or find me on X, LinkedIn, Medium, Dev.to, or the AWS Community.

For tons of great serverless content and discussions please join the Believe In Serverless community we have put together at this link: Believe In Serverless Community

Top comments (1)

Collapse
 
benbpyle profile image
Benjamen Pyle AWS Community Builders

Great article and thanks for sharing!