ResearchPad

AI-powered autonomous ML experimentation

Track experiments, surface insights, and debug models -- all from your editor.

Get Started View on GitHub

Quick Start

pip install researchpad
researchpad init
researchpad runserver

Then open http://localhost:8888 and start experimenting.

Features

Dashboard

A real-time overview of every experiment in your project. See status, metrics, and trends at a glance.

Experiments

Log each training run automatically. Compare hyperparameters, metrics, and outcomes across iterations.

Research

Capture literature reviews, hypotheses, and background research as structured artifacts tied to your experiments.

Debug

Analyze failing runs with structured debug reports. Pinpoint root causes and track fixes over time.

Insights

Accumulate learnings across experiments into a living knowledge base that grows with your project.

How It Works

ResearchPad follows an autonomous experiment loop:

1. Research -- Gather context, review literature, form hypotheses.
2. Experiment -- Run a training iteration and log results automatically.
3. Analyze -- The dashboard surfaces trends, regressions, and wins.
4. Debug -- When something goes wrong, generate structured debug reports.
5. Learn -- Insights accumulate so you never repeat a dead end.

All data lives in a local .researchpad/ directory inside your project -- no cloud accounts, no vendor lock-in.

graph LR
    A[Research] --> B[Experiment]
    B --> C[Analyze]
    C --> D{Success?}
    D -->|Yes| E[Record Insight]
    D -->|No| F[Debug]
    F --> A
    E --> A

Zero Dependencies

ResearchPad has no Python runtime dependencies. The CLI is pure Python; the UI is a self-contained Node.js server bundled into the package. Install it and go.

---

Ready to get started? Head to the Installation Guide.