Understand your data.

A white-box machine-learning environment for transparent, reproducible exploration.

Right where you left off, ready when you are.

Structured reasoning trail for projections, runs and analytical variations. Launch multiple models, compare projections side by side and resume any thread of analysis without retracing your steps.

Visual placeholder — Runs tree + multiple projection panels (the “Exploration Slices” view)
  • Cluster using K-Means, HDBSCAN, Agglomerative, Spectral, OPTICS, GMM and more.
  • Project with PCA, t-SNE, UMAP, Autoencoder, DEC, and more.
  • Visualize model outputs on a 2- or 3-d interactive plot.
  • Access runs and their variations with a lineage view — your reasoning history, visible, and organized.
  • Verify, analyze and compare by Silhouette, DBI, CHI, ARI, p-value, and more.

Clarity at every level — without the clutter.

Evaluate runs, clusters, and observations without losing your footing. Statistics, feature contributions and stability measures share one space, built for focus and interpretation.

Visual placeholder — three-image diagonal group (Run-vs-Run, Cluster Comparison, Sub-centroid / Observation)
  • Compare algorithms and parameters across methods.
  • Analyze clusters within one model — interpret feature contributions, stability, and statistics side by side.
  • Inspect neighborhoods and local density around cluster cores.
  • Focus on a single observation or patient to trace its relative position and attributes.
  • Track subjects over time for temporal insight [coming soon].
  • Predict future behavior [coming soon].

Keep your research reliable.

Audit-ready analyses stay consistent as your datasets evolve, maintaining reproducible and comparable insights over time. Collaborators can verify results without handling sensitive files, and snapshots preserve a visible trail from data to publication.

Visual placeholder — circular flow diagram: Dataset → Run → Share → Re-run → Dataset
  • Update datasets automatically using Google Drive, so your data is never stale.
  • Generate runs consistently, ensuring reproducible, verifiable results.
  • Share the exact computational lineage, not just the final figure.
  • Connect to other tools with one-click exports for datasets, projections, runs, and plots.
  • Publish interactive, immutable research artifacts using Snapshots for easy audit-ready viewing [coming soon].

No strings attached.

Start with the essentials — or supercharge your analyses with advanced models, GPU acceleration and more.
Plans start at $15 / month, cancel anytime.

Visual placeholder — modular grid: Core Environment + optional extensions (Deeper Models · More Compute · Shared Workspaces)
  • Utilize latent-space learning and advanced clustering algorithms.
  • Increase compute and storage quotas for larger datasets and worry-free exploration.
  • Unlock feature contributions, stability bootstrapping and more.
See the demo → Explore plans →

An evolving environment.

Ounias is expanding — new methods, better explanations, and deeper transparency. Check out what we're working on:

Snapshots

Make your analyses publication- and audit-ready.

Q1 2025

Supervised Methods

Train your own models with labeled data and supervised metrics alongside the existing unsupervised suite.

Q2 2025

AI-assisted Summaries

Use run outputs to bootstrap an editable report.

Q2 2025

Trajectory Predictions

Track and compare observations over time, enabling trajectories and improving model predictions.

Q2 2025

*Roadmap is indicative and adjusts with community feedback and emerging research directions.


Powered by Google Cloud — all datasets are stored securely behind Google's infrastructure. Users retain full ownership and can delete data at any time. Ounias never uses your data for analytics and does not share your data with any third party.

Read the Privacy Policy →