Speaker
Description
Ubuntu’s versatility as a platform for AI development is often underutilized, especially in local environments. This session will guide participants through setting up and experimenting with leading open-source LLMs (e.g., LLaMa 3.3, DeepSeek-R1, Qwen2.5-Max) on Ubuntu systems. We’ll explore practical use cases, from code generation to multilingual applications, while addressing hardware constraints and optimizing workflows for developers and enthusiasts in Asia. Attendees will gain actionable insights into integrating these tools into Ubuntu-based projects, fostering innovation in their communities
Content Summaries:
1. Introduction to Open-Source LLMs
2. Setting Up LLMs Locally on LXD (Ollama & Open WebUI)
3. Hands-On Workshop: Building Applications using VScode and LLM
4. Ethical and Community Considerations
Key Takeaways
- Practical skills to deploy and customize LLMs on Ubuntu (locally).
- Strategies to overcome hardware limitations in resource-constrained environments.
- Networking opportunities with Ubuntu’s AI/ML and localization communities in Asia
Summary
Ubuntu’s versatility as a platform for AI development is often underutilized, especially in local environments. This session will guide participants through setting up and experimenting with leading open-source LLMs (e.g., LLaMa 3.3, DeepSeek-R1, Qwen2.5-Max) on Ubuntu systems.
What audience can learn
- Practical skills to deploy and customize LLMs on Ubuntu.
- Strategies to overcome hardware limitations in resource-constrained environments.
- Networking opportunities with Ubuntu’s AI/ML and localization communities in Asia
Biography
Senior IT Consultant, Open Source Enthusiast, and long time contributor to Ubuntu Projects
Things to know or prepare for this session
- having Ubuntu Desktop natively or WSL on Windows (e.g 24.04 LTS) and dependencies (Python, CUDA etc)
- LXD platform
- Ollama and OpenWebUI platform
- VScode
- min. CPU/RAM requirements for running LLMs locally (e.g., 16GB RAM for 7B-parameter models) or cloud-based alternative with
- pre-downloaded smaller parameter models (small~7B-parameter or larger if have bigger hardware's specs )
| Difficulty level | Intermediate |
|---|