Explosion

Matthew Honnibal & Ines Montani are both forces of nature and massive benefactors of the open-source AI community so I can't wait to see what they'll build in the coming years. Also, it's a great practical example of how collaboration > competition in AI: Spacy & HF/transformers were kind of competitive libraries early on and we both decided to collaborate instead of competing for the benefit of the open-source AI community IMO. Long life Explosion! https://lnkd.in/eFTp7SCx

I want to take this opportunity to put the spot light on some wonderful (now ex-) colleagues who helped maintain spaCy and its open-source ecosystem for so many years. - First and foremost: Adriane B., the team's technical lead, ML engineer, resident computational linguist. She has fought all the Python packaging battles, trained spaCy models for 25+ languages, helped users on the forum and StackOverflow, and came up with efffiency tricks such as floret text embeddings (https://lnkd.in/e-FuYWgh). - Daniël de Kok and Madeeswaran Kannan, who have been responsible for countless efficiency improvements throughout the OS stack. This super productive duo also created 'curated-transformers' which supports transformer architectures built from the ground up. - Raphael Mitsch, a solid engineer who is equally good at scoping client projects as collaborating on Github to implement novel functionality. Raphael has been the driving force behind 'spacy-llm', allowing the integration of LLM prompting into structured spaCy pipelines. - Ákos Kádár, (PhD): A ML engineer with a strong academic background. Ákos cuts through the hype/bullshit to get to the core of a problem and its solution. He also implemented the experimental spaCy coref module, based on an end-to-end neural approach published in EMNLP 2021 (https://lnkd.in/eYziEP5w). - Peter Baumgartner: who helped us shape the consultancy efforts and implemented various tailored solutions for clients, succesfully marrying our open-source efforts with consulting (win-win!). I've had the honour to work with many more wonderful colleagues, including Basile Dura (now a freelance AI & Data consultant), Edward Schmuhl and Victoria Slocum (now at Weaviate), Lester James Miranda (currently at AI2), Paul McCann (https://lnkd.in/epGCi_EU), Richard Paul Hudson (author of coreferee) and Vinit Ravishankar. Then there's all the other Explosion (ex-) colleagues who worked on Prodigy (Teams), admin or customer success. Some of them have already moved on to the next challenge, others have taken a small break and are available right now if you're looking to hire exceptional NLP experts / ML engineers. If you're unsure who to reach out to but have some job openings that could be relevant - feel free to shoot me a message. Myself - I'm continuing with my one-woman consulting company OxyKodit, while also being actively engaged with open-source maintenance, such as Sebastián Ramírez Montaño's library Typer. Finally, I want to thank Matthew Honnibal & Ines Montani for having given me the opportunity to work on my two biggest passions for so many years - NLP and open-source, and for giving me the opportunity to lead the OS team even though probably none of us really knew what we were doing. Nevertheless, I'm so proud of everything we've accomplished. I wish Explosion the best of luck going forward, as I'm 100% convinced that spaCy and Prodigy should, can and will survive the current LLM-hype.

Company update: We're going back to our roots! We're back to running Explosion as a smaller, independent-minded and self-sufficient company. spaCy and Prodigy will stay stable and sustainable and we'll keep updating our stack with the latest technologies, without changing its core identity or purpose 💙 https://lnkd.in/evAaa4pn Thank you so much for all your support! Matthew Honnibal also wrote a more detailed blog post to share more background and some personal reflections: https://lnkd.in/gkRaQ5ad

⚗️ Develop small and performant models using LLMs Large language models (LLMs) work really well out of the box ✨ and as a result they are excellent tools for prototyping solutions and features that require AI. On the other hand, in production, you need reliable, auditable and cost effective workflows, something LLMs are not that great yet 👎 To make things worse, transitioning from an end to end, LLM based solution, to a production workflow with multiple components is not trivial since it requires a significant shift in not only the design of the solution but also the frameworks involved 🤯 To combat the above, you can architect your solution in a more production like fashion with the same components and use the same frameworks to prototype using an LLM and deploy your smaller modular models. An excellent example of this way of thinking is spacy and spacy-llm which allows you to quickly build an information extraction pipeline using your LLM of choice while allowing you to correct that data using humans and train much smaller more performant models at the same time 🚀 You can think of this process as distilling the knowledge from LLMs and humans into a compact model 🌟 🔗 Read more in this excellent blog from Explosion https://lnkd.in/e82Efdti

🌟 We’re super excited to have Ines Montani join us as a keynote speaker at #PDAmsterdam2024! Ines is ready to take us on an epic adventure through the world of Applied NLP in the age of Generative AI. 🚀 What’s on the Agenda? Get ready for a deep dive with Ines into the realm of Large Language Models (LLMs) and in-context learning, where the magic words are “prompts are all you need.” While prototyping has gotten a heck of a lot easier, making the leap to production can still be a wild ride. Ines will be sharing the secret spells—uh, lessons—learned from the front lines of real-world information extraction battles. Plus, she’ll reveal some game-changing strategies for building NLP systems that are not only smart but also tough enough for the real world. 👩💻 Meet Ines Montani: Ines isn’t just the co-founder and CEO of Explosion; she’s also a core developer of spaCy, a popular open-source library for Natural Language Processing in Python, and Prodigy, a modern annotation tool for creating training data for machine learning models. 🎟️ Why You Can’t Miss This: Join us to geek out over data science and snag some sage advice from Ines and other Data and AI wizards this September at PyData Amsterdam 2024. Grab your tickets now and prepare for an unforgettable experience! 🔗 https://lnkd.in/ek6FaNgh Can’t wait to see you there for some serious NLP fun and groundbreaking insights! #NLP #AI #MachineLearning #PyData #DataScience #GenerativeAI #LLMs #spaCy #Prodigy

In our latest episode, we're excited to host Ines Montani, a developer specializing in AI and NLP technology. She is the cofounder and CEO of Explosion and a co-developer of spaCy, the leading open-source library for NLP tasks in Python, as well as Prodigy, a cutting-edge annotation tool for creating training data for machine learning models. During our discussion, we explore: https://lnkd.in/g5XtvW4p ➡️ The focus areas for spaCy and how it all comes together ➡️ Applying generative AI technologies to specific industry problems ➡️ Common pain points in the field ➡️ Challenges faced by developers ➡️ The future of generative AI ➡️ An exciting Rapid-Fire Tune in now for a conversation that blends leadership wisdom, technological innovation, and practical advice. Kunal Jain #LeadingWithData #Explosion #spacy #ines #DataScience #AI #Podcast 

Check out my latest InfoQ w/ Ines Montani about the transformative power of open-source in AI! From democratizing tech to creating task-specific models, discover how open-source is shaping the future of AI and ensuring transparency, privacy, and innovation. #AI #OpenSource Read more here: https://lnkd.in/e477irds

10 years ago today Matthew Honnibal pushed the first commit to spaCy 🎉 Since then, the library has evolved as the field moved forward, but also stayed true to its core mission: industrial-strength #NLP and bringing structure to unstructured text. It's not always been easy building an open-source company while staying as independent and self-sufficient as possible and without compromising on our vision. There was a lot of trial and error and exploring new paths. I'm incredibly grateful for the supportive open-source community, the amazing team that helped us work on the library over the years and developers across so many different industries putting their trust in our stack and building on top of it. Thank you all 💙

This week for Once a Maintainer we spoke with NLP expert Sofie Van Landeghem, core maintainer of the popular open source NLP library spaCy. Sofie shared her thoughts on the evolution of NLP as applied to text mining for industry, how spaCy manages its roadmap, and why dependency management in python is so difficult. https://lnkd.in/eW55VZbe

⚗️ New blog post: A practical guide to human-in-the-loop distillation How to use state-of-the-art #LLMs in real-world applications and distill their knowledge into smaller and faster components you can run and maintain in-house. Summary below 👇 https://lnkd.in/eKAASN9H 1️⃣ Software in industry It’s important to keep in mind that AI development is still software development. We need modular, transparent, explainable, data-private, reliable & affordable solutions. This challenges black-box models and third-party APIs. But we can change that! 2️⃣ Strategies for applied #NLP Many real-world problems need structured data & knowledge about language and the world. LLMs are very good at this. But in-context learning doesn't mean transfer learning is obsolete – it's very competitive! Even more so if we control the data. 3️⃣ LLMs for predictive tasks With in-context learning, we start with a prompt template & corresponding parser to turn raw text into task-specific structured data. We can then use that output as training data with a human in the loop to correct errors and do better than the LLM. 4️⃣ Close gap between prototype and production Many projects face the prototype plateau and never make it to production. This is often a workflow problem. We should standardize inputs and outputs, have a robust evaluation, assess utility (not just accuracy) and work iteratively. 5️⃣ Structured data spacy-llm provides LLM-powered components for tasks like NER or relation extraction and the same structured output during prototyping and production – whether you ship the LLM pipeline or replace components with distilled models. https://lnkd.in/e4mKebgH 6️⃣ Human-in-the-loop distillation The goal is to create gold-standard data by extracting only the information we’re interested in, until the accuracy of the task-specific model exceeds the zero-/few-shot LLM baseline. Prodigy can help streamline this! https://prodi.gy 7️⃣ Case studies 1. https://lnkd.in/e4F_p4id 2. https://lnkd.in/ecBCC2Gh 8️⃣ Think of it as a refactoring process Just like with software, we need to break down larger problems and balance trade-offs between complexity and quality. We should also reassess dependencies: Can we replace larger models? Can we move a dependency from runtime to development? 9️⃣ Making problems easier You're allowed to make problems easier! This isn't university or a contest to solve the most difficult problems. Less operational complexity means less can go wrong. But getting there isn't trivial and a skill that develops through experience. 🔄 Conclusion In many real-world applications, AI models are going to be one function in a larger system. Refactoring & iteration is important. The right tools can get you past the prototype plateau. And there's no need to compromise on development best practices & data privacy!