Sphinx

Sphinx builds frontier AI agents for data science, combining statistical algorithms, multimodal understanding and business logic so that AI can interface with data directly. The company's systems are designed to accelerate the discovery of value from data at scale, covering data from small arrays of floats up to multi-terabyte tables. Its technical work spans AI agents, representation learning, agentic learning, multimodal understanding, statistical algorithms, machine learning and integration with data engineering and data infrastructure.

Three components form the core of the offering. Sphinx Copilot is an AI copilot that integrates into data science IDEs, including Jupyter and VSCode, as well as compute layers, databases and data lakes, augmenting every step of the data science process from data discovery and exploratory analysis to modeling and data-driven storytelling. A representation learning engine converts data from small arrays of floats to multi-terabyte tables into context that AI can ingest and understand, and an agentic learning system has agents explore and exploit across data modalities to find answers that maximise business value.

The company is backed by Y Combinator. Its founding team has deep roots at MIT and years of experience in AI research, and the company describes a research-oriented culture focused on frontier AI agents, with particular emphasis on the trustworthiness and reliability of AI tool outputs for enterprise use. Sphinx operates across enterprise data, data science, and analytics and business intelligence.