Over the past years, my career has been split across several domains that, at first glance, looked separate:

• Neural Networks and Deep Learning research
• Backend and distributed systems engineering
• AWS infrastructure and platform development
• Production systems, observability, logs, metrics, and reliability engineering

Recently, I realized these areas are converging.

Large Language Models are no longer just research projects or experimental tools. They are becoming infrastructure.

And I strongly believe the next big shift in software development will be the move toward local and self-hosted AI systems.

Companies increasingly want:

• control over their data,
• predictable infrastructure costs,
• private AI workflows,
• on-premise deployments,
• deeper customization,
• and AI integrated directly into internal systems.

The “one API for everything” model won’t fit every business.

That’s why I’m starting to focus on helping teams set up and integrate local/self-hosted LLM solutions - combining AI expertise with real-world software and infrastructure engineering experience.

This includes:

• deployment and orchestration,
• cloud/on-prem infrastructure,
• observability and monitoring,
• backend integrations,
• vector databases and retrieval,
• workflow automation,
• and building reliable AI systems that can actually run in production.

We’re still early. Many tools are rough around the edges. But that’s usually where the most important shifts begin.