Eloy Martinez
Engineer, builder, talker. 5th engineer at 🐂 Oxen.ai, where I've worked on blazing fast version control, fine-tuning pipelines, AI interfaces, custom deployment pipelines, and GTM. Previously shipped products at 🛠️ Mirego for clients like Familiprix, Airthings, and Archibald. Founder of 🦷 dentalbyaronin.com.
Do you like local LLMs? 🤖 (I do). Do you love pirates? 🏴☠️ (I really do).
Arr matey, click here and run a pirate LLM on your browser 🤯
Real AI running entirely in your browser. No servers involved. Warning: it's small and impressionable, so go easy on the lil guy.
Writing
Claude is Cool, but Your Brain is Way Cooler
AI energy consumption vs human cognition, why brains still win on efficiency
Gemma 4 Fine-Tuning and Deployments: A War Story
Technical deep dive into what it took to support Gemma 4 fine-tuning and deployments
Year of the Bold
Trying to apply it to life everyday
How We Cut Inference Costs from $46K to $7.5K
6x cost reduction fine-tuning Qwen-Image-Edit
How a $1 Qwen3-VL Fine-Tune Beat Gemini 3
Open-source VLM outperforming a flagship closed model
When to Fine-Tune an Image Model
Practical decision framework for fine-tuning vs. prompting
Oxen's Model Report (series)
Monthly analysis of new model releases and benchmarks
Speaking
Projects
Now
Based in Los Angeles 🌴 since Jan 2026. Trying to write more and talk less (you'll see why if we meet in person). Always curious, always building.
Beyond code (hover is my favorite css property)
I speak
Lived in
Been to
try the remote, it works ;)