Agent UX: Building developer tools for humans and their coding agentsFor a long time, visual SaaS apps have been the go-to for professional tools. With coding agents, this is changing – and the apps we build now need to work just as well for the agents operating them.
Happy 10th Anniversary!Exactly 10 years ago today, we officially founded Explosion. So much has happened since and we got to witness the field evolve, “AI” become mainstream, and hypes come and go. Thanks to all of you for your support over the past 10 years and beyond!
A quick chat with Ines & MattPy4AIImpromptu video interview on what inspired our talk and the conference vibes in sunny Pavia.
Vibe NLP for Applied NLPPyCon DE & PyDataWhat if we could take learnings from AI-powered coding agents and apply them to solving real-world NLP problems? In this talk, I’ll show how we’ve built a powerful virtual NLP assistant to help developers create practical and modular solutions that are small, fast and fully data-private.
Style tips for less experienced developers coding with AIGetting good performance out of LLM coding is less about prompts or agents, and more about the code you steer the model towards. Matt shares some lessons software engineers have learned over the years about building bigger things.
How AI is reshaping IT skillsconnect professionalGerman article featuring Ines’ take on the impact of AI on future-proof skills for IT professionals.
E^2GraphRAG: Streamlining Graph-based RAG for High Efficiency and EffectivenessZhao, Zhu, Guo, He, Li (2025)Instead of using LLMs for entity extraction, we employ the traditional NLP tool spaCy to extract entities, and use their co-occurrence in a chunk as relations.
Conquering PDFs: document understanding beyond plain textPyCon DE & PyDataIn this talk, Ines presents a new and modular approach for building robust document understanding systems, using state-of-the-art models and the awesome Python ecosystem.
How to advocate for modular NLP in the age of Generative AIWith all the hype around Generative AI, many are led to believe it’s the solution to everything. So how can you, as a developer, communicate the nuances and advocate for new and modular solutions that are better, easier and cheaper?
📚 spacy-layout v0.0.12Mar 8, 2025Support processing PDFs with context, add document index tables and more docs
Prozessvisualisierung mit generativer KI im PraxistestiX Magazin / HeiseGerman article by Nils Durner on visualizing technical processes with Generative AI, featuring spaCy and Presidio for PII anonymization.
Best Way to OCR a PDF in Python Python Tutorials for Digital HumanitiesTutorial by WJB Mattingly on how to use the new spaCy Layout package and Docling to convert PDFs to text.
PyLadies entrepreneurs and career developmentPyLadiesConPanel discussion about career challenges and starting your own business with Cheuk Ting Ho, Tereza Iofciu, Anwesha Das, Una Galyeva and Ines.
Distill Your LLMs and Surpass Their PerformanceInfoQ MagazineIn her presentation at InfoQ Dev Summit, Ines Montani provided the audience with practical solutions for using the latest state-of-the-art models in real-world applications and distilling their knowledge into smaller and faster components.
✨ prodigy v1.16.0Oct 22, 2024Modal plugin for on-demand deployment, cross-platform wheels and UI fixes
Applied NLP with LLMs: Beyond Black-Box MonolithsPyBerlinIn this talk, Ines shows some practical solutions for using the latest state-of-the-art models in real-world applications and distilling their knowledge into smaller and faster components.
Introducing Ellf: The missing piece for building NLP with coding assistantsEllf is our new tool that plugs into the coding assistant you already use, like Claude Code, Codex or GitHub Copilot, and gives it the missing pieces to develop custom AI systems.
Vibe NLP for Applied NLPBuild with AIWhat if we could take learnings from AI-powered coding agents and apply them to solving real-world NLP problems? In this talk, Ines shows you these new agentic workflows and how we’ve built a powerful virtual assistant to create practical solutions that are small, fast and data-private.
Taking back control of your AI developmentFeminist AI (PyData London)Short unconference presentation by Ines on the current focus of our work: empowering developers to build AI in-house again.
A Century of Immigration Rhetoric in the UK ParliamentGennaro, Vissens, Thornewill von Essen, Ravalde, Egan, Dave, Dable, Facini (2026)With all annotated snippets, we trained two transformer-based classifiers, each using a spaCy pipeline consisting of a RoBERTa base transformer model and spaCy’s textcat (text categorisation) component.
RiCoRecA: rich cooking recipe annotation schemaVentirozos, Jacob-Romero, Alrdahi, Clinch, Batista-Navarro (2026)The annotation process consists of two sections. Firstly, the annotator utilized a customized Prodigy interface to complete the NER and RC annotation tasks.
Sovereign AI systems instead of black box solutionsit-dailyGerman article featuring Ines’ take on AI in industry, the role of open source, and using Generative AI to create systems.
Developer Trends in 2025TalkPython PodcastDiscussion with Michael Kennedy, Calvin Hendryx-Parker, Gina Häußge, Richard Campbell and Ines.
Keyword Extraction, and Aspect Classification in Sinhala, English, and Code-Mixed ContentRizvi, Navojith, Adhikari, Senevirathna, Kasthurirathna, Abeywardhana (2025)Keyword extraction in English is performed with a hybrid approach comprising a fine-tuned spaCy NER model, FinBERT-based KeyBERT embeddings, YAKE, and EmbedRank, which results in a combined accuracy of 91.2%.
How Love Without Sound helps the music industry recover millions in revenue for artists with NLP, spaCy and ProdigyA case study on Love Without Sound’s innovative AI-powered tools for the music industry and law firms specializing in royalty negotiations.
✨ prodigy v1.18.0Feb 24, 2025Text editing during NER and span annotation, custom translations and more JavaScript features
What the history of the web can teach us about the future of AIHow will AI development look in the future? There is a lot we can learn from another groundbreaking technology: the web. This blog post takes a look at what the history of the web can teach us, and what this means for developers, models, open source and regulation.
Streaming spaCyJoin spaCy author and core developer Matt as he works on the library, develops features and fixes bugs, while chatting about all things NLP and open source. Every Thursday at 2pm CET and Friday at 11am CET.
Recognising non-named spatial entities in literary texts: a novel spatial entities classifierKababgi, Grisot, Pennino, Herrmann (2024)In this paper, we present a case study on the prediction of what we call ‘non-named spatial entities’ (NNSE) in a historical corpus of Swiss-German novels using a deep learning model in conjunction with BERT and Prodigy.
✨ prodigy v1.17.0Nov 18, 2024Pages UI for multi-page tasks like longer documents, PDFs or collections of images
Serverless custom NLP with LLMs, Modal and ProdigyIn this blog post, we’ll show you how you can go from an idea and little data to a fully custom information extraction model using Prodigy and Modal, no infrastructure or GPU setup required.
The 100 who are shaping AI in EuropeInes is featured among the top 100 individuals who are shaping Artificial Intelligence in Europe, compiled by French newspaper l’Opinion.
Ellf: Your virtual NLP engineerIn this new video series, Ines gives you an overview of our new tool and platform Ellf, its philosophy and how to use it to build your own custom NLP pipelines with the help of agents.
Build-to-own AI: Agentic Development for HumansPyData PyCon Yerevan KeynoteInstead of using LLM APIs, we can take back control and build our own systems, bootstrapped by agents. In this talk, Ines shows you these new agentic workflows, what they mean for developer tools and why code and open-source are more important than ever.
Show Us Your (Agent) SkillsPyMC Labs × Vanishing GradientsWhat are people at the top of the game building with AI agents and how are they doing it? In this episode, 6 developers are sharing their workflows.
How rightwing rhetoric has risen sharply in the UK parliamentThe GuardianIn-depth analysis by the Guardian in collaboration with University College London, powered by Prodigy for semi-automated data annotation and spaCy for classification model training and prediction.
Engineering a human-aligned LLM evaluation workflow with Prodigy and DSPyThis post demonstrates a human-in-the-loop workflow for developing and evaluating LLMs, using Prodigy and DSPy to create task-specific, human-aligned metrics that guide model optimization beyond generic evaluation measures.
Conquering PDFs: document understanding beyond plain textPyData LondonIn this talk, Ines presents a new and modular approach for building robust document understanding systems, using state-of-the-art models and the awesome Python ecosystem.
AI in Reality Fireside Chat: Enterprise AI & Open-Source InnovationPyCon DE & PyDataPanel discussion with Alexander CS Hendorf, Dr. Alexander Beck, Walid Mehanna and Ines.
KI ohne Ketten: Warum Open Source gegen Big Tech gewinnen kannUNMUTE IT Podcast (German)Interview with Ines on open source, LLMs, ethics and sustainable AI development.
Künstliche Intelligenz: Technologie der Zukunft – und warum Open Source die Karten neu mischtHeise KI-Woche 2025 (German)German talk on the future of Artificial Intelligence and the impact of open-source software and models.
What the history of the web can teach us about the future of AIPyCon+Web KeynoteIn this talk, Ines takes a look at what the history of the web can teach us about the future of AI, and what this means for developers, models, open source and regulation.
Prodigy Dashboard PluginThe new dashboard plugin adds a web application for managing annotations, data analytics and annotation progress, and is now available for early beta testing.
Cracking the Code: How to Start a Career in AIWelcome to the JungleShort video interview with Ines about the 4 skills job hunters can cultivate for a career in artificial intelligence.
From PDFs to AI-ready structured data: a deep diveThis blog post presents a new modular workflow for converting PDFs and similar documents to structured data and shows you how to build end-to-end document understanding and information extraction pipelines for industry use cases.
Accelerate your Career with Open-Source AIdotAIPanel discussion about making a career out of open-source software, featuring Gael Varoquaux (scikit-learn), Steeve Morin (ZML) and Ines.
Ellf: our new tool for agentic NLP developmentEllf gives your coding assistant like Claude Code superpowers for solving NLP problems and a platform to execute data-centric projects, automated annotation and more.
Atomic NLPAn applied NLP methodology inspired by Atomic Design: building reliable language understanding systems out of small, composable components instead of one big model and a prompt.
Building Multimodal Corpora Using Microtask Pipelines and Local AnnotatorsHotti, Vázquez, Jokipohja, Kalliokoski, Paakki, Suviranta, Hiippala (2026)To create the infrastructure needed for supporting this effort, we repurpose an existing commercial annotation tool, Prodigy, which we then enhance with additional components for combining the annotation tasks into pipelines, cross-validating the annotations and supporting annotator access to tasks.
The ultimate guide to optimizing annotation workflowsThis blog post collects tips and advice for how to build efficient human-in-the-loop data development workflows, break down business problems into actionable annotation steps and make the most of automation and model assistance.
Building AI with AIPyCon Ireland KeynoteAI-powered coding assistants have transformed the way we build software, and AI itself. In this talk, Ines shows why we should use LLMs to build systems instead of as systems, and why code is more important than ever, not less.
Applied NLP in the Age of Generative AI: Future-Proof Strategies for Banking and FinanceECONDAT KeynoteA modern approach and mindset for building future-proof NLP pipelines in-house, focusing on use cases from banking, finance and economics.
Feminist AI LAN PartyPyCon DE & PyDataThree days of workshops, hacking, creating, publishing and connecting locally, featuring a data development workshop with Prodigy and a session on hacking LLMs.
KI zwischen Freiheit und Kontrolle: The AI Revolution Will Not Be Monopolizeddata:unplugged (German)How should we envision the use of AI in practice? And are we heading further into a black box era with larger and larger models, obscured behind APIs controlled by big tech monopolies?
Mastering spaCyDéborah Mesquita, Duygu Altinok (Packt Publishing, 2025)Build structured NLP solutions with custom components and models powered by LLMs. By end of the book you will be empowered to build robust NLP pipelines and integrate them with web applications to build end-to-end solutions.
Using natural language processing to identify emergency department patients with incidental lung nodules requiring follow-upMoore, Socrates, Hesami, Denkewicz, Cavallo, Venkatesh, Taylor (2025)CT reports were annotated by MD raters using Prodigy software to develop a stepwise NLP “pipeline” that first excluded prior or known malignancy, determined the presence of a lung nodule, and then categorized any recommended follow-up. NLP was developed using a RoBERTa large language model on the spaCy platform.
spaCy Natural Language Processing: From Beginner to AdvancedGuan Wang, Xiaoquan Kong (2024)The first Chinese-language book on spaCy for beginners and experienced practitioners, covering traditional NLP techniques and how to leverage LLMs for various NLP tasks.
uOttawa at LegalLens-2024: Transformer-based Classification ExperimentsMeghdadi, Inkpen (2024)Our training utilizes the spaCy pipeline configured with a transformer model and a transition-based parser for NER tasks. The deberta-v3-base model has been selected for the main transformer architecture.
