The startup’s new platform merges full‑text, vector and SQL search on Parquet files, aiming to streamline how autonomous AI agents pull information.
Infino AI has closed a $7.5 million seed round to launch a unified data‑retrieval platform that lets autonomous AI agents query full‑text, vector and SQL indexes stored in Parquet files.
Unified Search Across Multiple Data Formats
The new service combines traditional SQL search with modern vector similarity and full‑text capabilities, eliminating the need for developers to stitch together separate tools when building AI‑driven applications.
By indexing Parquet files—a common storage format for large‑scale data warehouses—Infino AI enables agents to retrieve both structured and unstructured information with a single query interface.
How the Platform Works
Users upload their Parquet datasets to Infino’s cloud layer, where the system automatically generates three parallel indexes: a relational SQL index for exact matches, a vector index for semantic similarity, and a full‑text index for keyword searches.
An API call can then specify the desired search mode, or let the platform decide dynamically based on the query context, allowing autonomous agents to fetch the most relevant data without manual preprocessing.
Funding and Strategic Backers
The seed round was led by venture firms specializing in AI infrastructure, with participation from investors who have backed previous AI‑focused startups. The capital will be used to scale the platform, expand engineering resources, and add support for additional data formats.
- Enhance indexing performance for larger Parquet volumes
- Integrate with major cloud storage providers
- Add monitoring and analytics dashboards for query usage
Infino AI’s approach addresses a growing bottleneck in the AI ecosystem: the fragmented nature of data retrieval tools that forces developers to write custom glue code for each data source.
Our goal is to make data as accessible to AI agents as the web is to browsers, simplifying the retrieval pipeline from day one.
The company plans to roll out a beta program later this year, inviting early adopters to test the platform’s multi‑modal search capabilities in real‑world AI workloads.