International Conference
CECAR (Virtual Session), Sincelejo, Colombia
Lyra: From Thinking About AI to Integrating It Into Your Operating System
Overview
This talk was presented during the IX International Symposium on Systems Engineering, an international academic event focused on artificial intelligence, data science and software engineering that brought together students, researchers and professionals from multiple countries. :contentReference[oaicite:1]{index=1}
Speaking to an audience of approximately 200 attendees, I introduced the first public version of Lyra, an AI assistant for GNU/Linux designed to execute locally while using operating system context to provide more accurate and privacy-preserving assistance.
Rather than presenting AI as another cloud service, the session explored how specialized assistants can leverage local information to improve performance, reduce infrastructure costs and keep users in control of their own data.
Context
Most discussions around Artificial Intelligence focus on increasingly larger foundation models hosted in the cloud.
However, many operating system tasks do not require massive models or expensive infrastructure.
Lyra was created to explore a different approach based on:
- Domain-specific AI.
- Local execution.
- Operating system awareness.
- Vendor-independent architecture.
- Privacy-first design.
The presentation demonstrated how these ideas can be transformed into a practical software system.
Topics Covered
Why Build Local AI?
The talk examined the limitations of relying exclusively on cloud-hosted AI services.
Among the topics discussed were:
- Inference costs.
- Latency.
- Third-party dependencies.
- Privacy concerns.
- Limited access to operating system context.
It also explained why many system-level tasks can be solved efficiently using Small Language Models running locally.
Lyra’s Architecture
One of the main goals of the presentation was to explain the architecture behind the first version of Lyra.
The session covered components responsible for:
- AI orchestration.
- Context generation.
- Linux integration.
- Modular architecture.
- Backend and UI separation.
- Extensible software design.
Attendees were shown how these components collaborate to generate contextual responses about the operating system.
Optimizing AI for Local Execution
Another important topic focused on reducing infrastructure requirements without sacrificing usability.
Areas discussed included:
- Small Language Models (SLMs).
- Domain-specific context.
- Efficient inference.
- Lightweight architectures.
- Lower operational costs.
The presentation demonstrated that thoughtful software architecture is often as important as model size.
Engineering Challenges
The talk concluded by discussing several engineering challenges encountered while building Lyra.
These included:
- Safely extracting operating system information.
- Generating high-quality contextual prompts.
- Designing a modular architecture.
- Remaining vendor-independent.
- Building an extensible Open Source platform.
Key Takeaways
Some of the main ideas shared during the conference included:
- Bigger AI models are not always better.
- Local system context can significantly improve AI responses.
- Specialized assistants are often more efficient than general-purpose chatbots.
- Privacy can become a competitive engineering advantage.
- Good software architecture remains just as important as the language model itself.
Skills Demonstrated
This presentation showcases experience in:
- Artificial Intelligence
- AI Engineering
- Linux Development
- Software Architecture
- Small Language Models
- Local AI
- Open Source
- Technical Communication
- Systems Design
Why This Talk Matters
This presentation marked one of the first public technical introductions of Lyra at an international academic conference.
Beyond showcasing the project itself, it presented an alternative vision for AI software development—one where operating system context, efficient language models and privacy-first engineering work together to build practical intelligent assistants.
It reflects both my experience building AI-powered software and my interest in making artificial intelligence more efficient, transparent and accessible.
About the Event
The IX International Symposium on Systems Engineering brought together national and international speakers to discuss advances in artificial intelligence, data science, software engineering and digital transformation.
My session, “Lyra: From Thinking About AI to Integrating It Into Your Operating System”, was part of the official conference program and delivered virtually to symposium participants. :contentReference[oaicite:2]{index=2}
Frequently Asked Questions
What is Lyra?
Lyra is an AI assistant for GNU/Linux that combines Small Language Models with local operating system context to provide contextual assistance while reducing dependence on cloud-based AI services.
What was the goal of this presentation?
To demonstrate how AI assistants can be designed around local execution, operating system awareness and efficient software architecture instead of relying exclusively on cloud-hosted foundation models.
Who attended this talk?
Students, researchers, faculty members and software engineering professionals participating in the IX International Symposium on Systems Engineering.
Which version of Lyra was presented?
The conference showcased the first functional version of the project and explained the architectural decisions behind its implementation.
Why is Local AI important?
Local AI reduces infrastructure costs, lowers latency, improves privacy and enables intelligent systems to directly leverage information available on the user’s operating system.
Where can I find information about the event?
The official symposium program is available at: