MongoDB User Group Lisbon (MUG) @ Cloudflare

## Join us for a special collaborative edition of MongoDB User Group Lisbon hosted at Cloudflare!
This event brings together backend engineers, AI practitioners, cloud architects, and tech enthusiasts for an evening of cutting-edge talks, practical insights, and community networking.

🎤 Confirmed speakers:

• Dmytro Samoilov – Lead Android Engineer at Rollo.
  Presentation Topic: Hybrid AI: LLM on-device + Cloudflare Workers AI + MongoDB Atlas.
• Otavio Santana – Award-Winning Software Engineer, Java Champion, and Executive Committee Member for Java & Jakarta EE.
  Presentation Topic: Modern Software Architecture in the AI Era: New Patterns for Intelligent Platforms.

🗓️ Event Details:

• Data: 9th June
• Time: 18h- 20:30
• 📍Location: Av. da Índia 8 10, 1300-299 Lisboa.
  In Allo | Alcântara Lisbon Offices

🍕 Food & Drinks Provided
VERY IMPORTANT: For security reasons, only registered people in the event will be allowed to enter.
💬 Take this opportunity to meet amazing professionals, exchange ideas, and learn more about MongoDB and Gemini in the evolving world of Generative AI.
*** 👉 Register now — it’s FREE ***
💼 LinkedIn: https://www.linkedin.com/company/mongodb-user-group-lisbon
Are you interested in being a speaker? Fill the form and we will reach out https://forms.gle/ANpvY9Stp79qonfj7
Let’s build. Let’s share. Let’s grow — together. 💥
------------------------
Speaker: Dmytro Samoilov - Lead Android Engineer at Rollo.
Language: English
Title: Hybrid AI: LLM on-device + Cloudflare Workers AI + MongoDB Atlas
LinkedIn: [https://www.linkedin.com/in/samoilov-dmytro/](Dmytro Samoilov)
Abstract:
Every AI feature in your Android app starts the same way - an API call to the cloud. It works until your user base grows and the bill grows with it. Or until your users are on a train with no signal.
On-device LLMs like Gemma 4 can handle a surprising amount of work locally - transcription cleanup, entity extraction, text classification - at zero API cost. But most Android devices can't run them yet. Flagship phones with 12-16 GB of RAM can. The mid-range phone your average user carries? Not yet.

In this talk I'll demo a journaling app built on a three-tier hybrid architecture. On capable devices, Gemma 4 runs locally for transcription and text processing - fast, free, offline-capable. Cloudflare Workers AI handles summarization and embedding generation at the edge, and serves as the full AI backend for devices that can't run models locally. MongoDB Atlas stores entries with vector embeddings, enabling semantic search combined with structured filters across the user's history - something you can't easily do with a vector-only database.
The app dynamically routes between on-device and cloud based on what the phone can handle. Same user experience either way.

I'll show the routing logic, run both paths live on stage, and share real cost and latency benchmarks from actual devices. You'll leave knowing how to decide what runs locally versus in the cloud for your own app.

Bio:
Dmytro Samoilov - Android engineer with 8 years of experience building production apps for 500K+ users. Currently focused on on-device AI and hybrid cloud architectures for Android. Runs benchmarks on real devices, not marketing slides.

Duration: 30 minutes
----
Speaker: Otavio Santana – Award-Winning Software Engineer, Java Champion, and Executive Committee Member for Java & Jakarta EE.
Language: English
Title: Modern Software Architecture in the AI Era: New Patterns for Intelligent Platforms.
LinkedIn: [https://www.linkedin.com/in/otaviojava/](Otavio Santana)
Abstract:
Modern software architecture is entering a new era shaped by AI, intelligent systems, and probabilistic workflows. AI is not replacing software engineering or architecture—it is making them even more important. While AI helps teams write code faster, it does not automatically solve scalability, integration, governance, domain complexity, or long-term maintainability.
In this session, we will explore how architects and engineers can responsibly introduce AI into enterprise systems using patterns such as RAG, agents, semantic search, vector databases, embeddings, and AI-driven workflows. We will also examine how core architectural principles like modularity, observability, distributed systems, and boundaries evolve in AI-native applications.
Combining architectural thinking with practical implementation, this talk will show how Java and MongoDB Atlas can support intelligent search, enterprise integration, developer experience, and scalable AI-ready data architectures.
Duration: 30 minutes