Cognitive provides an in-database AI engine that adds deterministic inference to lakehouse SQL queries, allowing structured tables and transformed unstructured files to be reasoned over without separate vector search or RAG pipelines. The platform creates ontology views from S3 data, supports AI‑enhanced joins and classifications, and delivers low‑latency, token‑efficient results at petabyte scale.
Funding
Funding not disclosed
Founders
Product
Problem
Enterprises struggle to extract actionable insights from large volumes of structured and unstructured data stored in lakehouse environments because traditional approaches rely on separate vector databases, RAG pipelines, and external LLM orchestration, which introduce latency, probabilistic results, and complex integration overhead.
Solution
Cognitive offers an online cognitive processing platform that embeds deterministic AI inference directly into lakehouse databases via its Cognitive SQL engine. By transforming files in S3 and open formats into queryable ontology tables, users can run AI‑enhanced SQL statements that reason across relational data and document contents without separate vector search or RAG steps. The system uses smart brute‑force inference to deliver low‑latency, token‑efficient results that are deterministic and reusable across queries. Persistent inference results are cached and incrementally refreshed, enabling millisecond‑scale query responses at petabyte scale. This approach simplifies deployment, reduces operational complexity, and provides consistent, accurate insights for finance, healthcare, sales, and talent analytics use cases.
Target Audience
Primary customers are data engineering and analytics teams in enterprises—particularly in finance, healthcare, sales, and talent management—that need to query and reason over large mixed‑type datasets directly within their lakehouse environments.
Features
- In‑database Cognitive SQL engine that performs deterministic inference on both structured tables and transformed unstructured files
- Ontology Views that automatically convert PDFs, docs, spreadsheets, and multimedia stored in S3 into queryable tables without ETL or migration
- Cognitive Join capability to reason across relational data and document content in a single SQL statement
- Cognitive Classify functions for semantic classification of rows and documents at scale (e.g., fraud risk, complaint type)
- Cognitive Topic Discovery to extract emerging topics from existing datasets such as clinical trial reports
- Token‑efficient, low‑latency query execution with incremental background refresh of inference results
- Seamless integration with existing lakehouse architectures via standard SQL interfaces and open data formats