AI Enthusiast • Independent Developer • Classfied ops
Exploring local AI systems, privacy-first infrastructure, and experimental tools ☕🥣
I’m a developer who spends a lot of time experimenting with local AI, sovereign infrastructure, and tools that run entirely under the user's control.
My background is actually in classified, which probably explains why I approach software the way I do — systems thinking, diagnostics, and a bias toward building tools that work reliably in real-world conditions.
Most of my projects start as late-night experiments after work — usually fueled by curiosity, coffee, and a desire to understand how things work under the hood. Sometimes those experiments grow into real systems.
One of those experiments became Somi Project.
SOMI is an open-source framework designed to run privacy-first, locally-hosted AI agents and operational intelligence systems.
The idea is simple:
Your AI should run for you, on your machine, under your control.
Key goals:
- Fully offline / self-hosted capable
- Designed for data sovereignty and privacy
- Modular architecture for agents, tools, memory, and workflows
- Compatible with modern local LLM ecosystems
- Built as a research playground for agentic systems and long-context reasoning
The project is heavily inspired by the emerging sovereign AI movement — where individuals and organizations operate AI systems locally rather than through centralized platforms.
Status: Actively evolving (currently in the v0.9x range). The architecture changes often because the project doubles as a learning laboratory for new models, memory systems, and tool orchestration patterns.
Recent improvements include:
- Expanded web and research integration
- Better model compatibility with modern local LLMs
- Internal architectural cleanup and modularization
- Early groundwork for autonomous agent workflows
Repo: https://github.com/Somi-Project/Somi
Project updates: https://x.com/SomiProject
Stars, issues, ideas, and technical discussions are always welcome.
Areas I enjoy exploring:
AI Systems
- Local LLM deployment (Qwen, Llama, etc.)
- Agent frameworks and orchestration
- Long-context reasoning
- AI memory architectures
Software
- Python (primary language)
- Rust / Go experimentation for performance-critical pieces
- System design and modular architectures
Infrastructure
- Self-hosted environments
- Docker and containerized AI stacks
- Offline-first systems
Philosophy
- Minimize SaaS dependence
- Build systems that remain useful even without internet access
- Keep things transparent and hackable
- Most development happens late at night with coffee nearby
- I enjoy understanding systems from first principles
- My projects are usually experiments that spiral into full frameworks
- I tend to build tools I wish existed for my own workflows
- Occasionally assisted by a curious cat who likes sitting on keyboards
X / Twitter https://x.com/frostyflak3s
Where I usually share:
- AI papers and experiments
- updates on SomiProject
- thoughts on local AI and sovereign systems
Feel free to:
- star repositories
- open issues
- suggest improvements
- contribute ideas
I especially enjoy discussions around local AI architectures and experimental agent systems.