In this hands-on, fast-paced workshop, participants will spend the day building real LLM-powered applications through an immersive hackathon. After a short crash course covering the essential tools, frameworks, and design patterns for modern LLM systems, teams will collaborate to prototype innovative solutions, applying concepts like retrieval, fine-tuning, and agent workflows directly in code. Mentors will guide participants throughout, with mini-presentations and theory bites woven into the day to reinforce key ideas. Designed for developers with Python experience and some familiarity with LLM implementation, this event balances learning with creativity, collaboration, and fun—culminating in project demos, social connections, and prizes for standout teams announced live at the conference.

We will kick off the hackathon with a ~1 hour presentation of the format and core concepts to know. Then, we'll let you off to form teams and start coding directly! The guidelines will be made in a way that makes it realistic to arrive to a final deployed application within a day so that, by the end of the day, each team (one person from the team) will present their final work and "pitch" it to a jury in under 5 minutes. The jury will then deliberate to find a winner that will be announced during the main conference to win a prize!

We'll then have mini crash courses for those that want to listen on while others can keep coding:

- APIs, FastAPI, MCP...
- Vibe coding
- Orchestrating agents
- Evaluating LLM pipelines
- Building quick UIs with tools like Gradio or streamlit

‍Requirement: Basic coding skills in Python. Ideally already have implemented LLMs through APIs or locally to be able to start a project quickly (or join a team with someone that can help you).

Target Audience: Anyone who wants to practice implementing LLMs in real applications.