Making Patent Data Accessible and Reusable

Yesterday I released my first open source project for the intellectual property community. I'd like to tell you a little more about it, how you can use it, and why you might want to.    First — what's a CLI? It stands for "command-line interface." Think of it as a tool your computer (or your AI agent) can run by typing a command, instead of clicking through a website. So what can you actually do with it?  - Search every patent a company has ever filed or been granted  - Look up any patent application and see its full prosecution history  - Read the actual claims, abstract, and full specification text of any granted patent  - Download every document in a patent's file history — office actions, amendments, drawings, all of it  - Trace a patent family tree across continuations, divisionals, and CIPs  - See who owns a patent and every time it's been assigned or transferred  - Check if a patent has been challenged at the Patent Trial and Appeal Board  - Look up appeal and interference decisions  - Search petition decisions  - Export any of the above to a spreadsheet with one command  - Download the USPTO's weekly bulk data dumps — patent grants, applications, file wrappers  - Build a complete due diligence profile of any application with a single command Basically: if it's in the USPTO's public data, you can get to it. Do you need to learn how to use the command line? Nope. Point any AI coding agent, Claude Code, Cowork, Codex, OpenCode, Claw at the GitHub link in the comments. It will read the documentation, walk you through installing the tool, help you get your API key, and have you up and running in 15 minutes. Max. The project is open source. The API key is free. Your agent will have full access to the USPTO's data. It runs locally on your machine, only connects to the USPTO. The AI agent is all on your terms. Enjoy.

🚀 Introducing AI Agent-Powered Patent Intelligence Framework for Strategic IP Decision Support 📄 Patent documents are among the richest sources of innovation insight—but they’re often the least accessible. This whitepaper presents a breakthrough framework that transforms static legal disclosures into dynamic, AI-interpreted, and market-aligned intelligence. Built using modular AI agents, the system parses, interprets, and maps patent claims to real-world applications—bridging legal scope, technical architecture, and commercial opportunity. 🔍 What’s inside: ✅ A structured 5-stage agent framework for claims-to-market intelligence ✅ Case studies in sleep tech, CRISPR, and digital therapeutics ✅ Cross-sector mapping to healthcare, telecom, consumer electronics, and automotive ✅ Real-time outputs: claim charts, product-feature matrices, and licensing strategies ✅ Use cases for FTO, M&A, licensing, litigation, innovation scouting, IP valuation, and more 💡 From “What is protected?” to “Where does it win?”—this framework helps IP attorneys, R&D leaders, investors, and licensing teams turn complex patents into clear, actionable decisions. #AIAgents #IntellectualProperty #PatentStrategy #Innovation #Licensing #IPManagement

🤓 New paper alert: open #dataset on scientific citations in USPTO Office Actions 👩💻 Most research that uses patent citation data relies on front-page citations. But those data can be systematically incomplete in ways that matter for inference: ❌ Front-page "examiner" citations may actually be applicant-submitted references simply retained by the examiner — you can't tell who truly found what. ❌ IDS submissions are not fully reflected on the front page — so "applicant citations" are not the full disclosure set. ❌ Abandoned/refused applications have no front-page citations at all — a major selection issue. ❌ Even for granted patents, front-page citations exclude prior art that drove rejections and claim amendments during prosecution. ➡️ Office Action citations address some of these shortcomings: they exist for abandoned applications, cover the examination record, and capture what examiners actually communicated during prosecution. What we release (open data + code): ✅ ~850,000 Office Action citations, classified into 14 reference types; ✅ ~265,000 citations to scientific literature, parsed/cleaned/disambiguated, and linked to OpenAlex; ✅ A fully documented pipeline (including consolidation via Crossref when needed). This project was developed in collaboration with Google, jointly with Kyle Higham, Emma Scharfmann, Steve Gong, and Hannah Kotula 🙏 If you work on examiner behavior, science–technology linkages, or citation-based metrics, I hope you'll download the data and build on it. Paper (DOI): https://lnkd.in/eSvRF3nT Data (Figshare): https://lnkd.in/eTKM2Sud | https://lnkd.in/enT562-8 Code (GitHub): https://lnkd.in/eDQFm65b #patents #innovation #intellectualproperty #USPTO #opendata #datascience #scienceofscience #economics

Unlocking Federal Innovation: Industry Licensing of Government IP The federal government holds one of the most underutilized innovation assets in America: thousands of patented technologies available for licensing from the DoD, VA, NIH, and other federal entities. Platforms every innovator should know 𝗖𝗿𝗼𝘀𝘀-𝗔𝗴𝗲𝗻𝗰𝘆 𝗣𝗹𝗮𝘁𝗳𝗼𝗿𝗺𝘀 • 𝗠𝗲𝗱𝗶𝗰𝗮𝗹 𝗧𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝘆 𝗘𝗻𝘁𝗲𝗿𝗽𝗿𝗶𝘀𝗲 𝗖𝗼𝗻𝘀𝗼𝗿𝘁𝗶𝘂𝗺 has search portal focused on medical and health-related innovations at that is a valuable starting point for discovering cutting-edge patents and tech that could accelerate your product development pipeline (MTEC; https://lnkd.in/g2i2phZi). • 𝗗𝗢𝗘 𝗟𝗮𝗯 𝗣𝗮𝗿𝘁𝗻𝗲𝗿𝗶𝗻𝗴 𝗦𝗲𝗿𝘃𝗶𝗰𝗲 centralizes access to all 17 national laboratories with 25,000+ patents covering energy, materials, computing, and advanced manufacturing (labpartnering.org). • 𝗙𝗲𝗱𝗲𝗿𝗮𝗹 𝗟𝗮𝗯𝗼𝗿𝗮𝘁𝗼𝗿𝘆 𝗖𝗼𝗻𝘀𝗼𝗿𝘁𝗶𝘂𝗺 connects you to 300+ federal labs across all agencies. Their FLC Business search tool provides free access to tech, and their Technology Locator Service offers personalized assistance at zero cost (FLC; federallabs.org; https://lnkd.in/gDUemvS8). • 𝗧𝗲𝗰𝗵𝗟𝗶𝗻𝗸 is the official intermediary for DoD and VA, managing 60% of all Defense Department licenses. Their database includes 9,000+ active DoD and VA patents. All are searchable and available to license with expert support (https://techlinkcenter.org). 𝗔𝗴𝗲𝗻𝗰𝘆-𝗦𝗽𝗲𝗰𝗶𝗳𝗶𝗰 • 𝗡𝗔𝗦𝗔 offers exceptionally startup-friendly terms: no upfront fees for 3 years under their Startup NASA License, with 1,200+ aerospace and robotics patents available (thttps://lnkd.in/gqj2mkwW). • 𝗡𝗜𝗛 𝗧𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝘆 𝗧𝗿𝗮𝗻𝘀𝗳𝗲𝗿 manages the largest biomedical portfolio with 4,446 patents, including technologies behind 34 FDA-approved drugs that generated $2B in royalties (techtransfer.nih.gov). 𝗗𝗶𝗿𝗲𝗰𝘁 𝗣𝗮𝘁𝗲𝗻𝘁 𝗦𝗲𝗮𝗿𝗰𝗵𝗶𝗻𝗴 • 𝗨𝗦𝗣𝗧𝗢 𝗣𝗮𝘁𝗲𝗻𝘁 𝗣𝘂𝗯𝗹𝗶𝗰 𝗦𝗲𝗮𝗿𝗰𝗵 provides comprehensive access to all federal patents. Search by assignee: "United States of America," "Department of Energy," "NASA," or "Secretary of the Navy" (https://lnkd.in/gZVjQHqy). • 𝗚𝗼𝗼𝗴𝗹𝗲 𝗣𝗮𝘁𝗲𝗻𝘁𝘀 offers user-friendly searching with advanced filters for agency names and technology classifications (patents.google.com). Annually, Federal labs generate 1,600+ patented inventions from publicly-funded research. This tech is available under favorable licensing terms that offer lower fees, startup-friendly provisions, and expert support throughout the process. This list is not exhaustive. Some additional links are in the comments. Have you explored federal technology transfer for your organization? What's been your experience with government technology licensing? #TechTransfer  #IP  #Licensing  #MTEC #MILMEDConnect

Have you ever tried to search through millions of black-and-white technical patent drawings using only text 🥵 ? Through the European Patent Office Academic Research Programme (ARP), we collaborated with TIB – Leibniz-Informationszentrum Technik und Naturwissenschaften und Universitätsbibliothek Hannover to develop an alternative. By adapting large vision-language models to the unique domain of patents, we are moving beyond text searches and projecting images and text into a unified embedding space. In the linked article, I explore how this unlocks incredible new ways to query prior art: 🔍 Targeted Subpart Search (cropping a specific component to find precise matches) ➕ Multimodal Queries (combining an image with text, like "+ A human heart") ✍️ Sketch and Search (drawing a modification directly onto an image to retrieve specific configurations) This is another great step forward for our philosophy: AI doing the heavy lifting with the human expert in the driver's seat. A huge thank you to the ARP team from TIB @Sushil Awale, Eric Müller-Budack and Dr. Ralph Ewerth and our EPO Data Science colleagues Rahim and Franco as well as Daniel Schneider, Head of Search tools, and many others for this fantastic collaboration. Read the full article below to see the visual examples and learn more about the architecture, and let me know your thoughts in the comments! 👇 #ArtificialIntelligence #DataScience