🤿 The last two days I had the pleasure of joining the AWS Deep Dive Days - Generative AI in our Munich office.

🤿 The last two days I had the pleasure of joining the AWS Deep Dive Days - Generative AI in our Munich office. The name of the event again, 3rd time already, hold the promise. Deep content from techies for techies. Loved it.    ⚠️ Like always the event summary is obviously biased by my preference, my other work commitments during the conference, appetite for networking and an element of randomness/going with the flow (key ingredients for broad explorations). Not possible to cover a 3+ track conference as a single person anyhow.    💚 Each day started with impressive keynotes. At the first day Slavik, like always, blew our minds with loads of interesting thoughts, views and analysis. If we only could have a full-day keynote slot for him to present all his content ;) - Emerson talked about the evolution of Alexa on the 2nd day’s keynote, including a very impressive demo of the upcoming Alexa+. Game Changer. Can’t wait to welcome Alexa+ in my home.

📰 Robert and Diana presented about t-online’s Article Extender project. The talk was packed with good insights into lessons learned so far, highlighted the need of observability and evaluation for high quality content in production and gave very practical insights how to implement this successfully in production. Took a lot of insights and ideas from the talk and the extensive Q&A session.    ⚽ Michael and Philipp dived deep in their session “How SAP Sports One and Amazon Bedrock Transform Game Preparation” on how they use Generative AI to take football match analysis to the next level. Impressive. Wanna build something with LangGraph Agents? This is for you1   😎 Olivier and Lorenzo gave a lot of practical insights in their session “Designing and Building Multi-Agent Systems: From Architecture to Implementation” on how to build those agents everyone is (just?) talking about. I guess the deep Q&A after the session continued in the networking area for a long while :)

🏝️ Close to my heart as efficiency will be key for all of us is the topic of model distillation / compression, which lets you shrink a model’s memory and compute footprint for inference while trading away a slice of the larger model’s accuracy (the classic reference is Hinton, Vinyals, and Dean’s 2015 “Distilling the Knowledge in a Neural Network”). My colleagues Mia and Jennifer gave an impressive presentation on this topic: “Five Practical Insights into Model Distillation”.   💬 And there have been many more session. Also, super interesting demos have been presented as part of the event. Kudos to Katja, Aman,Tobi and all the others who guided participants through this.    🙏 Big kudos to the organiser team, all the hidden champions and participants which made this event a truly special one!   ➡️ Overall, it was nice to see that the event kept a very good depth. Topics evolved with the GenAI evolution, but the content covered pretty much everything from current research topics to detailed lessons learned from production implementations.

➡️I took a lot of input away for my upcoming talk at the AWS Summit Hamburg. More details on this can be found in my earlier post ([2]).

Links below!

• AI
• AWS
• Events