Ollama Api

I started using Ollama from its first public release, immediately recognizing the potential of this technology for running Large Language Models in a local environment. This early adoption has allowed me to acquire deep knowledge of the API and its evolution, becoming a reference point for implementing self-hosted AI solutions in the projects I work on.

Code autocompletion with local models

I use models running through Ollama on a daily basis for code autocompletion during development. This choice is deliberate: in my work I often handle sensitive codebases that cannot be shared with cloud AI providers like OpenAI or Anthropic. Projects for parliamentary institutions, enterprise systems, and applications dealing with confidential data require an approach that guarantees complete source code confidentiality. By running models on my local infrastructure, I obtain the benefits of AI assistance while maintaining full control over data, without a single line of code leaving my development environment.

Dedicated infrastructure for local AI

I manage a home server room specifically configured for running artificial intelligence models. This infrastructure allows me to experiment with different models, optimize hardware configurations, and test solutions in realistic scenarios before proposing them in professional contexts. Hands-on experience in managing computational resources for AI workloads represents an added value that enables me to competently tackle projects requiring on-premise solutions.

RAG systems development

I have designed and implemented multiple Retrieval Augmented Generation systems that integrate Ollama as the inference engine. These systems combine the power of LLMs with semantic search over document bases, enabling contextualized and accurate responses from specific document corpora. I have made one of my implementations public in the Local RAG Example repository, a practical guide demonstrating how to build a local RAG system using Go, PostgreSQL with pgvector for vector indexing, and Ollama for response generation. This open source project represents my approach to sharing knowledge with the community.

Privacy and security as priorities

The choice to use Ollama reflects a broader awareness regarding cybersecurity in modern software development. In a context where AI is becoming increasingly integrated into workflows, the ability to keep sensitive data within one's own security perimeter is fundamental. This expertise makes me particularly suited to work on projects where code and data confidentiality is non-negotiable.