Tensorlake

When we joined Tensorlake, the product was initially built for engineers — an API-first platform intended to be deployed directly into user infrastructure. While this approach offered speed and flexibility, it became clear that we needed more robust user-facing tools to support scale, usability, and broader adoption.

We began by designing and building the Document Ingestion UI, evolving it from a basic upload interface into a powerful, flexible tool that supported:

Custom schema input for structured data extraction

Support for advanced features like signature detection

A schema builder UI so users could create extraction schemas visually instead of pasting raw code

Collection creation, allowing users to reuse configurations across large document sets

Following that, we focused on the Workflows interface — the frontend layer for managing serverless data workflows. This view made it easy to visualize:

The sequence of functions and tasks in a pipeline

The output of each task

Task status and errors during execution

We shipped these features through continuous iteration, grounded in real user feedback and tight cross-functional collaboration.

These product improvements significantly enhanced how customers interacted with Tensorlake and led directly to meaningful business outcomes:

Helped close $$$$+ in enterprise deals

Supported acquisition of 5–7 new high-value customers

Contributed to Tensorlake’s strong competitive positioning in the fast-growing unstructured data infra space