Explosion, NLP, Generative AI, Entrepreneurship

2023-10-26 · Source: Explosion · Developer tools and consulting for AI, Machine Learning and NLP - Explosion.ai · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Entrepreneurship & Start-ups · Depth: Intermediate, extended

Summary

Ines Montani, co-founder and CEO of Explosion, discusses her journey into Natural Language Processing and the evolution of spaCy, their flagship open-source library. She highlights spaCy's design philosophy, emphasizing an opinionated, production-ready approach that prioritizes accuracy, speed, and a unified API for real-world applications. Montani draws parallels between the development of web technologies and machine learning, stressing the need for deep understanding beyond automated tools. Explosion's strategy includes developing new components like "span-cat" and "coreference resolution", integrating large language models via spaCy-LLM, and enhancing developer experience with tools like config files and a project system. The company is also launching Prodigy Teams, a SaaS platform for collaborative, data-private ML workflows. Montani advocates for a startup philosophy focused on sustainability and creating tangible value, rather than solely pursuing rapid growth, and shares insights on team building with "tree-shaped" skills.

Key takeaway

For NLP Engineers and ML Directors building production systems, prioritize robust, opinionated tools like spaCy that offer a clear pipeline approach. Focus on distilling large language models into smaller, task-specific solutions for better control, efficiency, and data privacy, rather than relying solely on general-purpose LLM APIs. Consider adopting collaborative platforms like Prodigy Teams to streamline data labeling and model development, ensuring your solutions are both effective and sustainable.

Key insights

SpaCy provides an opinionated, production-ready NLP pipeline, emphasizing practical application, developer experience, and sustainable business models.

Principles

• Prioritize production-ready, efficient NLP tools.
• Business understanding drives effective ML solutions.
• Sustainability over rapid growth in software startups.

Method

SpaCy's pipeline approach combines rule-based, deep learning, and LLM-powered components for flexible, practical NLP, enabling rapid prototyping and distillation of large models into task-specific solutions.

In practice

• Use spaCy for robust, production-grade NLP pipelines.
• Distill LLM knowledge into smaller, faster models.
• Employ Prodigy Teams for collaborative data labeling.

Topics

• Natural Language Processing
• spaCy Library
• Production ML
• Large Language Models
• Open-Source Business
• Data Annotation

Best for: AI Engineer, NLP Engineer, Machine Learning Engineer, Director of AI/ML

Related on AIssential

• Building Industrial-Strength NLP Pipelines — Production-grade NLP requires pragmatic problem decomposition, efficient tools, and continuous data-model iteration.
• spacy-llm: From quick prototyping with LLMs to more reliable and efficient NLP solutions — spacy-llm integrates LLMs into spaCy for structured NLP, moving from rapid prototyping to reliable, cost-efficient solutions.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Explosion · Developer tools and consulting for AI, Machine Learning and NLP - Explosion.ai.