Oliver
Perrin
I’m a machine learning engineer on the founding team of a stealth startup, where I lead the ML technical direction across model development, evaluation, inference, and integration.
Outside that work, I build LiminalML and explore NLP, multi-task learning, and post-training. I also write about the models I build and the ideas behind them.
Read my writing
My work
Stealth startup
I’m the only ML engineer on the team. Starting from the CEO’s product ideas, I work out the ML approach and build it, along with the backend and data infrastructure it depends on. The company is pre-launch and most of this work is proprietary, so I keep the details private.
ExperienceLiminalML
I founded LiminalML to build software for learning and explaining machine learning. I design, build, and run the company myself.
LoRA update space
Change the adapter rank to see its outputs expand from a line to a plane to a volume.
A three-output toy model with fixed synthetic inputs. It shows the adapter correction, not trained model activations.
Open in StudioPreparing the interactive visual…
Writing
Implementation notes, explanations,
and things I learned the hard way.
LexiMind
My open-source research project, started as an undergraduate. With a custom PyTorch transformer, I’m testing whether merging task-specialist models beats joint training once you count the cost of training the specialists.
The repository also has a book-discovery app. Its content-based recommendations run separately from the research model.
Open source
PlayAxis
Find events, follow teams, log training, and connect with people who play.
React / FastAPISartheAtlas
Explore Le Mans cars and the circuit through interactive engineering studies and maps.
Three.js / MapLibrePulse
A local dashboard for coding-agent usage, subscription allowances, and reset times.
Electron / Node.js