Skip to main content
Welcome to Elata’s machine-learning stack for brain and physiological data. This repo is where you:
  • Ingest and organize raw signals (e.g. EEG and related streams)
  • Build, train, and evaluate ML models
  • Expose models via CLI tools and a web application
  • Share reproducible experiments through notebooks and versioned model artifacts
This page is your landing hub:
  • A quick mental model of the repo
  • How branches are typically used
  • Where to go next depending on what you’re trying to do
If you just want to run something locally right away, skip ahead to Pick your path.

​
What lives in this repo?

brain-ml is designed as a full stack ML workspace:
  • Core library (src/)
    Data loading, preprocessing, feature pipelines, model definitions, training/eval loops, and shared utilities.
  • Web app (webapp/)
    Frontend for visualizing runs, inspecting datasets, and interacting with trained models through a browser. This is where “non-ML” teammates usually live.
  • Notebooks (notebooks/)
    Research and exploration: quick experiments, ablation studies, visualizations, and one-off analyses.
  • Model artifacts (models/)
    Checkpoints, exported weights, and metadata. Typically gitignored; think of this as the “outputs” of your experiments.
  • Docs (docs/)
    Everything that powers the Mintlify docs site:
    • High-level guides under /docs/guides/
    • Tutorials under /docs/tutorials/
    • API and SDK reference under /docs/reference/
    • This page under /docs/getting-started/index.md
  • Scripts (scripts/)
    Helper scripts for documentation and tooling:
    • generate_python_docs.sh: builds Python API docs into docs/reference/sdk/python/autogenerated/
    • generate_typedoc.sh: builds frontend TypeScript docs into docs/reference/sdk/typescript/typedoc/
The root also includes standard project files like README.md, CHANGELOG.md, .github/workflows/ for CI, and .gitignore.

​
Branches: how work is usually organized

brain-ml is designed to support multiple parallel tracks of work: production-ready pipelines, active development, and more experimental research. The exact branch names may vary between teams and forks, but you’ll typically see three kinds of branches:
  1. Stable / release branches
    Long-lived branches that correspond to deployable, stable code (e.g. a main or release/* branch).
    • Intended for: production deployments, reproducible results, and published docs.
    • CI should be green; tests and core pipelines expected to pass.
  2. Integration / feature branches
    Branches where active development happens (e.g. dev, webapp/*, pipeline/*, feature/*).
    • Intended for: day-to-day work, integrating new models or data pipelines, webapp changes, etc.
    • May contain breaking changes, WIP experiments, or newer dependencies.
  3. Experiment / research branches
    Short-lived branches (often prefixed like exp/* or research/*) for model or dataset experiments.
    • Intended for: trying new architectures, loss functions, or preprocessing approaches.
    • Not guaranteed to be stable; may never be merged.
To see what’s available in your clone: 
# Clone the repo (SSH)
git clone git@github.com:Elata-Biosciences/brain-ml.git
cd brain-ml

# Fetch all remote branches
git fetch origin --prune

# List local and remote branches
git branch -a