In 2023, Devi and Jianna co-founded Leafpress (YC S23), an energy management platform that relied heavily on parsing utility bills. What seemed straightforward quickly spiraled into a nightmare. Bills arrived handwritten, annotated, or formatted inconsistently, causing traditional OCR tools to fail spectacularly. "Every line item mattered, and preserving the structure was critical—otherwise, account numbers wouldn’t match usage data, and things would fall apart," the founders recall. They tried every market solution, resorting to a clunky, human-in-the-loop pipeline that still delivered inaccuracies. This pain birthed Cardinal AI: an OCR engine built not in a lab, but from the trenches of real-world document chaos.

Cardinal tackles the Achilles' heel of conventional OCR—handling messy, unstructured documents—with AI-powered precision. It processes everything from medical records and contracts to low-quality scans and faxes, extracting data while maintaining contextual integrity. Key features include:

  • Structure Preservation: Accurately captures complex table layouts and cell relationships, crucial for financial or legal documents.
  • Multi-Format Mastery: Interprets handwritten notes, checkboxes, signatures, and even redlined text in contracts.
  • Barcode & Image Intelligence: Automatically identifies 1D/2D barcodes and extracts images with metadata.
  • Language Agnosticism: Processes and translates multi-language documents reliably.
  • Resilience: Handles poor-quality inputs like photocopies with high reliability, reducing preprocessing needs.

Security is core to Cardinal’s design, addressing enterprise concerns with rigorous, audited practices. Data remains encrypted in transit and at rest, compliant with healthcare standards like HIPAA, and can be auto-deleted post-processing or kept within a client’s infrastructure. For developers, this means fewer custom pipelines and more focus on high-value tasks, while industries like healthcare and legal gain a tool that mitigates compliance risks.

The founders’ journey underscores a broader industry gap: as AI advances, solutions must solve actual user pains, not just technical benchmarks. Cardinal’s launch invites developers to test its mettle—Devi and Jianna are actively seeking feedback at [email protected] and [email protected]. In a world drowning in unstructured data, tools like this aren’t just convenient; they’re the bedrock of scalable automation.

Source: Cardinal AI