Investigate systematically
Surface structure, predictive signals, transitions, uncertainty, and directions worth deeper investigation.
A no-code research analysis environment
Complex Data → Coherent Insight
Ounias helps researchers investigate tabular data across complementary methods, compare what they reveal, and preserve the evidence behind each finding—without coding or assembling an ML workflow.
No code required
Methods remain visible
Projects & run lineage
Reproducible findings
ProjectionPCA · 2D
Metric pillarsbalanced
Key findings3 signals
Three groups recur across methods.
Feature set A explains the strongest split.
One subgroup remains structurally unstable.
Cross-method finding
Three-group structure persists across 5 configurations
Keep the analysis in context
Organize datasets, runs, variations, artifacts, and findings without losing the history behind them.
Publish evidence with context
Turn selected findings into durable, interactive analytical records for publication, review, and audit.
Designed for research, not automatic conclusions
Ounias helps you investigate what the data supports.
It surfaces recurring patterns, contradictions, uncertainty, method sensitivity, and promising directions for deeper study. The researcher retains responsibility for the question, domain context, interpretation, and scientific judgment.
An analysis should make the path to a finding clearer—not conceal it behind an answer.
Ounias structures
The analytical process
- Method execution and guided configuration
- Cross-run and cross-method comparison
- Metrics, explanations, stability, and lineage
- Projects, artifacts, history, and provenance
The researcher provides
The scientific judgment
- The research question and domain context
- Interpretation of patterns and contradictions
- Decisions about relevance and significance
- Further validation and substantive conclusions
How Ounias works
From dataset to a structured investigation
Guided workflows remove the need to code or assemble an ML stack, while projects, runs, methods, assumptions, and supporting evidence remain visible throughout the investigation.
01
Upload & understand
Import a tabular dataset, define feature roles, validate its structure, and profile its quality, compatibility, and analytical characteristics.
or import from Drive
Dataset profile
Rows8,742
Features31
Missingness2.6%
Method fitHigh
Outliers4.1%
LongitudinalDetected
02
Explore across methods
Examine the data through clustering, supervised learning, dimensionality reduction, and complementary configurations—manually or through guided evaluation.
Analytical paths
Feature-space clustering
Embedded-space clustering
Supervised analysis
Longitudinal analysis
Evaluation plan
03
Organize the investigation
Keep datasets, runs, variations, artifacts, and findings organized within projects. Revisit any result and trace how it was produced without reconstructing the analysis from notebooks or filenames.
Analytical historyFeature-space investigation
Context preserved
KMeans · baselineOriginal run · 31 features
.84Feature-pruned variationDerived from baseline · 24 features
.87Filtered cohortDerived variation · 7,918 rows
.81HDBSCAN · UMAPAlternative method family
.78Stability assessmentArtifact attached · perturbation mode
stable04
Compare, explain & validate
Compare analytical alternatives, inspect projections and label characteristics, examine feature contributions and stability, and see where methods agree or diverge.
Feature contribution
Biomarker A
Age
Measure C
Score D
Feature E
Supporting findings
✓
Stable core82% of assignments persist under perturbation.
↗
Primary dividerTwo features explain most of the separation.
!
Boundary subgroupOne region remains method-sensitive.
05
Follow change over time
Analyze repeated observations, trace entity trajectories, measure transitions, and understand how labels and target values evolve over time.
Entity trajectories
State transitions
A → B
34%
B → C
26%
A → C
15%
C → B
10%
B → A
7%
06
Publish analytical context
Turn a selected analytical state into a fingerprinted, interactive record that preserves findings, evidence, configurations, and lineage for publication, review, or audit.
Included views
Findings & comparisons
Projection & filters
Insights & explanations
Stability assessment
Longitudinal analysis
Built to fit the research stack
Bring data in, investigate it, and carry the results forward
Ounias is designed as an analytical layer within the tools researchers already use, rather than a replacement for the rest of their workflow.
Bring data in
Connect research sources
Google Drive
Google Sheets
OneDrive · soon
Vertex AI · soon
Ounias
Investigate across methods
Profile, evaluate, compare, explain, and preserve analytical lineage in one reproducible workspace.
Compare
Explain
Preserve
Carry results forward
Keep the work usable elsewhere
MLflow
Interactive snapshots
Notion · soon
Published analytical context
Explore findings together with their evidence
A snapshot turns a selected analytical state into a durable, interactive record. Readers can inspect findings, supporting evidence, method comparisons, and provenance rather than receiving a detached conclusion.
- Move from a stated finding directly to the plots, metrics, and comparisons supporting it.
- Inspect the analytical decisions, included runs, and lineage that define the published context.
- Share or cite the analysis without publishing the underlying raw observations.
A snapshot presents an analysis with its supporting context—not a certified scientific conclusion.
Open Live SnapshotProjectionPrimary structure
FindingCross-method
Three stable cohorts recur across four method families.
A compact boundary subgroup requires closer review.
Evidence5 pillars
Visual placeholder — replace with the final demo snapshot capture.
Research & trust
Accessible to use. Transparent enough to defend.
Ounias removes the engineering burden without hiding the methods, assumptions, evidence, or analytical history behind a result.
Evidence across methods
Ounias compares complementary analytical approaches rather than treating one model or metric as the complete answer.
Reproducible by design
Seeds, configurations, dataset identity, artifacts, metrics, and run lineage are preserved together.
Accessible, not opaque
Guided defaults and a no-code workspace reduce technical setup while keeping methods, configurations, and outputs open to inspection.
Publishable analytical context
Snapshots preserve selected findings with their evidence, provenance, and citation details in a durable interactive record.
Analytical capabilities
One workspace, multiple perspectives
Ounias combines methods that are commonly fragmented across notebooks, libraries, and disconnected tools.
01 · Discover
Structural discovery
Identify meaningful organization in complex datasets and examine it in both original and learned spaces.
02 · Predict
Predictive analysis
Compare supervised methods, understand target relationships, and inspect the features driving model behavior.
03 · Validate
Reliability & comparison
Test whether findings persist, compare results through a shared framework, and retain every analytical branch.
04 · Follow
Longitudinal analysis
Study repeated observations as trajectories rather than isolated rows, revealing movement, transitions, and temporal structure.
Investigate without building the infrastructure
Start with the question. Explore what the data supports.
Bring the research question and domain judgment. Ounias handles the analytical workflow, comparison, organization, and provenance—without requiring code or ML engineering.