Pulse AI (YC S24)

We just launched Pulse STUDIO, a zero-shot, production-grade extraction model for documents and spreadsheets. (Playground, docs, and API access links in comments!) Why? There's a million OCR tools on the market. We spent the summer in the Y Combinator batch working directly with F100s and growth stage startups building supply chain optimization software. Problem: their invoices, purchase orders, and goods receipts contained tens of millions of pages that they had to parse through every day, and every existing solution was not end-to-end. Developers building RAG pipelines or running other ETL processes have to spend hours in the pre-processing stage using outdated parsing libraries, something we suffered through. Solution: integrate our fine tuned vision language model (VLMs) with more computer vision + classical ML classification algorithms. We’re just getting started – the model is getting faster, more efficient and more accurate every single day. Let us know what you think! To companies building in the voice + video space, Pulse STUDIO will soon support audio extraction with background noise filtering, speaker identification and separation, and transcription with timestamps. We will also be launching a novel spreadsheet reasoning tool built on top of Excel + Google Sheets using our models as the foundational layer. Stay tuned!

Pulse AI (YC S24) just launched an API for production-grade unstructured document extraction, turning complex information into LLM-ready inputs. No training required. Approximately 75% of enterprise data is unstructured, the majority of this is directly within PDF files. This makes it extremely difficult to build RAG applications with this data, and ingestion is often the bottleneck. The team tested every tool on the market and found they lacked accurate contextual understanding, multi-column PDFs, and multimodal documents. Most current technologies are simply wrappers on Textract or Gemini, which have their own inherent flaws. Pulse trained its own set of Vision Language Models (VLMs) and OCR techniques to bridge this gap. Pulse reached state-of-the-art (SOTA) performance on its vision model for documents and spreadsheets. The API processes all PDF types, including handwritten documents, foreign languages, and more. It seamlessly integrates into new or existing engineering workflows as well. They are also actively working on a novel reasoning tool on spreadsheets using this technology – stay tuned. Pulse's API saw initial success across supply chain teams in the three-way match process and is deployed in companies across hardware, healthcare, manufacturing industries, and more. Sign up at studio.trypulse.ai. Congrats on the launch Sid Manchkanti and Ritvik Pandey! ?? https://lnkd.in/gKW37rcH

We just launched NEXUS, an AI-powered search engine designed specifically for supply chain teams to unify their data. Supply chain professionals often face the challenge of sifting through massive amounts of ERP data—purchase orders, invoices, emails, and more—scattered across different systems. Finding the right information quickly is crucial, yet analysts waste ~20% of their time with traditional search tools. NEXUS changes that. Built with real-time vector embeddings, NEXUS seamlessly integrates data from ERPs, emails, Slack, and documents, creating a unified search experience. Our zero-shot vision model ensures that even complex, multi-page documents are embedded with their context preserved. NEXUS also uses a sophisticated re-ranking algorithm to solve the issue of close similarity scores, ensuring that the most relevant results always rise to the top. Whether you're looking for pending POs assigned to specific team members or tracking approvals with long lead times, NEXUS delivers precise, context-aware results instantly.

We just launched NEXUS, an AI-powered search engine designed specifically for supply chain teams to unify their data. Supply chain professionals often face the challenge of sifting through massive amounts of ERP data—purchase orders, invoices, emails, and more—scattered across different systems. Finding the right information quickly is crucial, yet analysts waste ~20% of their time with traditional search tools. NEXUS changes that. Built with real-time vector embeddings, NEXUS seamlessly integrates data from ERPs, emails, Slack, and documents, creating a unified search experience. Our zero-shot vision model ensures that even complex, multi-page documents are embedded with their context preserved. NEXUS also uses a sophisticated re-ranking algorithm to solve the issue of close similarity scores, ensuring that the most relevant results always rise to the top. Whether you're looking for pending POs assigned to specific team members or tracking approvals with long lead times, NEXUS delivers precise, context-aware results instantly.

Pulse AI (YC S24) is launching a new LLM-based tool to automatically perform the three-way matching process for procurement and accounting teams. Pulse AI, founded by Sid Manchkanti and Ritvik Pandey, previously launched a set of ML tools to automate procurement workflows and various ERP actions. Their fine-tuned models offer production-ready reliability and enable workflow automations across entire supply chain organizations. Three-way matching is a critical process in accounting and procurement teams that helps identify fraud and ensures the accuracy of all incoming invoices. Despite its significance, this procedure remains highly labor-intensive. Pulse AI has developed an in-house document vision model that extracts relevant pieces of information such as line items, delivery dates, supplier addresses, and tabular data. Using this structured output, they streamline the matching process, enhancing accuracy while significantly reducing human time spent. Congrats, Sid and Ritvik, on the launch! ?? https://lnkd.in/gteCc2N6

Pulse AI (YC S24) is an inference layer that automates and analyzes data from legacy supply chain systems with AI. Legacy ERPs are extremely antiquated, expensive, and slow. However, they are a vital part of nearly every company. Tedious and time-consuming manual work results in large inefficiencies inside of supply chain organizations, costing businesses billions of dollars annually. Current companies rely on Pulse AI to automate procurement workflows, supplier reliability analysis, and ML-based demand forecasting, giving them more time to prioritize critical and innovative tasks. Employees are saving hours a day and are ~25% more productive on previously tedious data transcription tasks. After working in ML at NVIDIA and Tesla, Sid Manchkanti and Ritvik Pandey — who also have extensive research backgrounds in machine learning from Berkeley and Georgia Tech— founded Pulse AI with the mission to revolutionize supply chain infrastructure with LLMs. Congrats on the launch, Sid and Ritvik — and welcome to YC! ?? https://lnkd.in/giThgHRx