Pinsight builds open source large language and vision language models specialized for data analytics and document understanding. Its TOLMA model analyzes tables to generate reports, charts, and maps, while DOLMA extracts structured JSON from scanned documents. The models can be self-hosted for privacy compliance and customized freely, unlike commercial alternatives.
Funding
Funding not disclosed
Founders
Product
Problem
General-purpose commercial AI models like ChatGPT and BARD are not optimized for specialized tasks such as structured data analysis or extracting information from scanned documents. Organizations handling sensitive data face compliance challenges when using cloud-based commercial models, and these tools lack the customization needed for domain-specific analytics workflows.
Solution
Pinsight provides open source large language models (LLMs) and vision language models (VLMs) purpose-built for data analytics and document understanding. The TOLMA model specializes in analyzing tabular data and generating reports, charts, and maps, while the DOLMA model converts scanned document images into structured JSON output. Because the models are open source, organizations can deploy them behind their own firewalls to maintain data privacy and meet regulatory requirements. This approach combines specialized performance with full control over deployment and customization.
Target Audience
Primary customers are data-driven organizations and enterprises that need specialized, privacy-compliant AI for analytics and document processing, including those in regulated industries such as healthcare, finance, and legal sectors.
Features
- TOLMA: state-of-the-art LLM for table analysis, report generation, chart creation, and map production
- DOLMA: multi-modal VLM that extracts structured JSON from scanned document images
- Open source models hosted publicly, free to use and customizable for specific needs
- Self-hosted deployment option for privacy-preserving environments
- Designed to meet GDPR, SOC2, and HIPAA compliance requirements
- Narrow, specialized expertise in analytics and document reading rather than generic assistance