Projects I work on personal projects focused on simplicity, performance, and learning by doing. Check out my repos for tools, scripts, and experiments in system optimization and security.

01 | ABOUT

Freelance software engineer with commercial experience delivering work for companies. I write systems-level code in C and C++, build machine learning tooling in Python and CUDA, and care about software that is small, readable and easy to audit.

My approach is simple: less code means fewer bugs, a smaller attack surface and a faster review. I prefer minimal dependencies, POSIX-compatible tooling and designs that one person can read and fully understand in an afternoon.

I also contribute to open source projects when I can be useful.

02 | FOCUS

• Systems programming: C and C++, POSIX, low-level tooling, performance work
• Machine learning engineering: model quantization, QLoRA fine-tuning, inference optimization
• GPU computing: CUDA kernels and GPU-accelerated pipelines
• Security and auditing: minimal, reviewable code with a small trusted base
• Alternative operating systems: Darwin, FreeBSD, OpenBSD, NetBSD, Plan 9, illumos
• Terminal-centric workflow: Neovim, tmux, shell tooling, suckless-style software

03 | LANGUAGES

Fluent in C, C++ and Python, with professional experience in eight more languages.
Also writing Holy C, Pascal and 1C.

04 | OPERATING SYSTEMS

I run and target a range of systems, with a particular interest in Darwin internals and the BSD family.

05 | MACHINE LEARNING

Hands-on engineering experience in applied AI and ML, with emphasis on efficient training and inference on constrained hardware.

• Quantization of large language models for lower memory use and faster inference
• QLoRA fine-tuning and other parameter-efficient methods
• CUDA development for custom GPU workloads
• Model publishing and maintenance on Hugging Face

Hugging Face: huggingface.co/tinyopsec

06 | TOOLING

Daily work happens in the terminal, with tmux, Make and suckless-style software.

07 | PRINCIPLES

1. Simplicity. The simplest design that solves the problem is the right one.
2. Auditability. Code should be short enough to read in full and clear enough to verify.
3. Minimalism. Every dependency and every feature has a cost, so each one must justify itself.
4. Performance. Measure first, then optimize what matters.
5. Learning by building. Real projects teach more than tutorials.

08 | WORK WITH ME

Available for freelance engagements in:

• Systems and low-level development in C and C++
• Machine learning engineering: quantization, fine-tuning, deployment
• CUDA and performance optimization
• Code review and security-minded refactoring
• Tooling and automation

Contact: open an issue or discussion on any repository, or reach out through the links on my GitHub profile.

09 | STATISTICS