Leap Labs offers Disco, an automated exploratory data analysis platform that fits neural networks to tabular datasets and applies interpretability methods to surface statistically significant patterns. It validates findings on hold-out data, contextualizes them with existing literature, and provides ranked results with p-values, effect sizes, and evidence. The platform is accessible via API, Python SDK, or MCP for integration into AI agent workflows.
Funding
Funding not disclosed
Founders
Product
Problem
Traditional exploratory data analysis relies on manual hypothesis testing, which is time-consuming, biased by prior assumptions, and prone to missing complex non-linear interactions, thresholds, and subgroup effects. Scientific literature and aggregated summaries are lossy abstractions that obscure valuable nuance present in raw data, limiting discovery.
Solution
Leap Labs provides Disco, an automated pattern discovery engine that fits neural networks to tabular data and applies interpretability methods to extract the patterns they learned. All findings are validated on hold-out data and contextualized with existing literature, producing a ranked list of statistically significant patterns with p-values, effect sizes, evidence, and context. Disco operates without requiring prior hypotheses, enabling superhuman exploratory data analysis. The platform is accessible via API, Python SDK, or MCP, making it suitable for AI agents and developers seeking structured, validated pattern discovery.
Target Audience
Primary users are data scientists, researchers, and AI developers who need to discover validated patterns in tabular datasets without manual hypothesis testing, as well as teams building AI agents that require automated pattern discovery capabilities.
Features
- Automated neural network fitting and interpretability extraction to uncover non-linear interactions, thresholds, and subgroup effects
- Statistical validation on hold-out data with p-values and effect sizes for each reported pattern
- Literature contextualization that links findings to existing scientific evidence
- Ranked output of novel findings with structured results and citations
- API, Python SDK, and MCP interfaces for seamless integration into agent workflows
- Free tier for public data with unlimited published analyses; paid plans for private data and deeper analysis