Builder's Night: Train Your Own Model

​You already ship features on top of LLMs. Come train a small language model of your own and leave with it deployed and running.

​Bring a task you currently send to a frontier model, such as classifying tickets, extracting structured fields from messy text, routing tool calls, or scoring outputs. Don’t have one in mind? Pick one of our starter tasks and just try the thing.

​The distil labs platform generates the training data and fine-tunes a small model using a much larger model as its teacher. Your first model will be trained and deployed in less than an hour, leaving the rest of the evening to evaluate, iterate, and see how far you can push it.

​You won’t need to touch a training loop, provision a GPU, or label a dataset. Bring a laptop with Claude Code installed and signed in. Bring API keys if you want to evaluate against something real.

​The CLI supports macOS and Linux. If you’re on Windows, we’ll have a REST API route ready for you.

​Pizza, drinks, and the engineers who built the platform will be in the room all evening.