Pulse cofounder Ritvik Pandey has stated his plans to integrate the Gemini 2.0 artificial intelligence model alongside other foundation models for document processing. He stated in a LinkedIn post that Pulse was building a custom solution that combined traditional computer vision algorithms with vision transformers alongside the latest LLMs.
Pulse (which stands for Product-Grade Unstructured Document Extraction) is a Y Combinator-backed document processing engine. It makes use of the document vision model to accurately parse documents and create LLM-ready texts.
READ: Y Combinator startup launch: Unlock AI-driven user assistance with Layup (December 10, 2024)
According to Ritvik Pandey–who had previously worked for Goldman Sachs and Tesla–research has revealed several problems with document processing. LLMs process images through lossy transformations, optimizing for semantic meaning while discarding crucial visual details , which can be especially problematic for financial data and tables.The probabilistic nature of token prediction means also LLMs often make subtle, dangerous substitutions that appear plausible but can be entirely wrong (think “rn” vs “m” or decimal point shifts in financial data). Pulse is trying to find a solution to this.
READ: Parse.ly founders raise $6.4 million for new AI startup (December 5, 2024)
Gemini 2.0, Google’s latest AI model, was launched on Wednesday. The suite of models includes 2.0 Flash, which is billed as a “workhorse model, optimal for high-volume, high-frequency tasks at scale,” as well as 2.0 Pro Experimental for coding performance, and 2.0 Flash-Lite, which the company calls its “most cost-efficient model yet.”Gemini Flash costs developers 10 cents per million tokens for text, image and video inputs, while Flash-Lite, its more cost-effective version, costs 0.75 of a cent for the same. Tokens refer to each individual unit of data that the model processes. These releases are part of Google’s strategy of investing in AI agents, as competition over AI heats up between companies.