Upgrade to Pro — share decks privately, control downloads, hide ads and more …

Applied NLP in the Age of Generative AI: Future...

Applied NLP in the Age of Generative AI: Future-Proof Strategies for Banking and Finance

Large Language Models (LLMs) and in-context learning have introduced a new paradigm for developing natural language understanding systems: prompts are all you need! Prototyping has never been easier, but not all prototypes give a smooth path to production. Many new ideas that are emerging also challenge existing workflows in industry that require modularity, transparency and data privacy. In this talk, I'll share the most important lessons we've learned from solving real-world information extraction problems in industry, and show you a new approach and mindset for building modular and future-proof NLP pipelines in-house.

Avatar for Ines Montani

Ines Montani

June 05, 2025
Tweet

Resources

How S&P Global is making markets more transparent with NLP, spaCy and Prodigy

https://explosion.ai/blog/sp-global-commodities

A case study on S&P Global’s efficient information extraction pipelines for real-time commodities trading insights in a high-security environment using human-in-the-loop distillation.

The Window-Knocking Machine Test

https://ines.io/blog/window-knocking-machine-test/

How will technology shape our world going forward? And what tools and products should we build? When imagining what the future could look like, it helps to look back in time and compare past visions to our reality today.

What the history of the web can teach us about the future of AI

https://explosion.ai/blog/history-web-future-ai

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.

A practical guide to human-in-the-loop distillation

https://explosion.ai/blog/human-in-the-loop-distillation

This blog post presents practical solutions for using the latest state-of-the-art models in real-world applications and distilling their knowledge into smaller and faster components that you can run and maintain in-house.

From PDFs to AI-ready structured data: a deep dive

https://explosion.ai/blog/pdfs-nlp-structured-data

How to build end-to-end document understanding and information extraction pipelines for industry use cases.

Using LLMs for human-in-the-loop distillation in Prodigy

https://prodi.gy/docs/large-language-models

Prodigy comes with preconfigured workflows for using LLMs to speed up and automate annotation and create datasets for distilling large generative models into more accurate, smaller, faster and fully private task-specific components.

More Decks by Ines Montani

Other Decks in Technology

Transcript

  1. Generative AI can write spaCy code! Open-source library for industrial-

    strength natural language processing spacy.io 390m+ downloads
  2. Modern scriptable annotation tool for machine learning developers prodigy.ai Alex

    Smith Developer Kim Miller Analyst GPT-4 API 900+ companies 10k+ users
  3. MIXTRAL GPT-4 good contextual results ⚠ transparency ⚠ e iciency

    easy to use & configure fast prototyping LLM
  4. MIXTRAL GPT-4 good contextual results ⚠ data privacy ⚠ transparency

    ⚠ e iciency easy to use & configure fast prototyping LLM
  5. Are you designing a window-knocking machine or an alarm clock?

    “knocker-uppers” ines.io/blog/window-knocking-machine-test
  6. Hello, I ’ m Toni ’ s virtual assistant and

    I help schedule appointments. Do you have time at 1pm on Monday? No, but Tuesday would work for me. Okay, please confirm: Tuesday at 1pm? 1pm is unideal but 3pm would work. Toni doesn ’ t have availability at 3pm but I could offer a slot at 4pm or 5 : 30pm. Which time zone is this by the way? I ’ m in CET. ines.io/blog/window-knocking-machine-test
  7. Hello, I ’ m Toni ’ s virtual assistant and

    I help schedule appointments. Do you have time at 1pm on Monday? No, but Tuesday would work for me. Okay, please confirm: Tuesday at 1pm? 1pm is unideal but 3pm would work. Toni doesn ’ t have availability at 3pm but I could offer a slot at 4pm or 5 : 30pm. Which time zone is this by the way? I ’ m in CET. Calendly ines.io/blog/window-knocking-machine-test
  8. Hello, I ’ m Toni ’ s virtual assistant and

    I help schedule appointments. Do you have time at 1pm on Monday? No, but Tuesday would work for me. Okay, please confirm: Tuesday at 1pm? 1pm is unideal but 3pm would work. Toni doesn ’ t have availability at 3pm but I could offer a slot at 4pm or 5 : 30pm. Which time zone is this by the way? I ’ m in CET. Calendly “window-knocking machine” “alarm clock” ines.io/blog/window-knocking-machine-test
  9. ines.io/blog/window-knocking-machine-test What ’ s the total services revenue from 2023?

    $2,923,531 How many clients is that in total? 29 ⏺ ⏺ ⏺
  10. ines.io/blog/window-knocking-machine-test What ’ s the total services revenue from 2023?

    $2,923,531 How many clients is that in total? 29 ⏺ ⏺ ⏺ 🔮 LLM 📚 database 🤖 agents ⚙ query Retrieval-Augmented Generation
  11. ines.io/blog/window-knocking-machine-test 2023 Year Services Type ACME Inc. FooBar GmbH NLPCorp

    XKCD Ltd. Python AG 432,032 82,000 1,500 193,000 91,320 $ 2,625,032 Clients (28) Revenue What ’ s the total services revenue from 2023? $2,923,531 How many clients is that in total? 29 ⏺ ⏺ ⏺ 🔮 LLM 📚 database 🤖 agents ⚙ query Retrieval-Augmented Generation
  12. ines.io/blog/window-knocking-machine-test 2023 Year Services Type ACME Inc. FooBar GmbH NLPCorp

    XKCD Ltd. Python AG 432,032 82,000 1,500 193,000 91,320 $ 2,625,032 Clients (28) Revenue AI still needs product decisions! What ’ s the total services revenue from 2023? $2,923,531 How many clients is that in total? 29 ⏺ ⏺ ⏺ 🔮 LLM 📚 database 🤖 agents ⚙ query Retrieval-Augmented Generation
  13. dynamic pages static pages compile static data at build time

    explosion.ai/blog/history-web-future-ai static pages WEB
  14. dynamic pages static pages compile static data at build time

    custom models AI explosion.ai/blog/history-web-future-ai static pages WEB
  15. dynamic pages static pages pretrained models compile static data at

    build time custom models AI explosion.ai/blog/history-web-future-ai static pages WEB
  16. dynamic pages static pages custom models pretrained models compile static

    data at build time custom models AI explosion.ai/blog/history-web-future-ai static pages WEB
  17. dynamic pages static pages custom models distill models into smaller,

    faster and private components pretrained models compile static data at build time custom models AI explosion.ai/blog/history-web-future-ai static pages WEB
  18. explosion.ai/blog/sp-global-commodities 99% F-score 6mb model size 16k+ words/second • S&P

    Global: real-time commodities trading insights by extracting structured attributes
  19. explosion.ai/blog/sp-global-commodities 99% F-score 6mb model size 16k+ words/second • S&P

    Global: real-time commodities trading insights by extracting structured attributes • high-security environment
  20. explosion.ai/blog/sp-global-commodities 99% F-score 6mb model size 16k+ words/second • S&P

    Global: real-time commodities trading insights by extracting structured attributes • high-security environment • 10× data development speedup with humans and LLM in the loop
  21. explosion.ai/blog/sp-global-commodities 99% F-score 6mb model size 16k+ words/second • S&P

    Global: real-time commodities trading insights by extracting structured attributes • high-security environment • 10× data development speedup with humans and LLM in the loop • 8+ market pipelines in production
  22. explosion.ai/blog/sp-global-commodities 99% F-score 6mb model size 16k+ words/second • S&P

    Global: real-time commodities trading insights by extracting structured attributes • high-security environment • 10× data development speedup with humans and LLM in the loop • 8+ market pipelines in production
  23. explosion.ai/blog/sp-global-commodities 🧑💻 human experts in the loop 🚀 structured data

    📦 model package 📚 task-specific data ➕ 🔮 suggestions from LLM
  24. 🧑💻 developer tooling 🧡 open source 👩🔬 subject matter experts

    🔄 iteration 🤖 development support 🛠 refactoring
  25. 🧑💻 developer tooling 🧡 open source 👩🔬 subject matter experts

    🔄 iteration 🤖 development support 🛠 refactoring 🏢 in-house development
  26. 🧑💻 developer tooling 🧡 open source 👩🔬 subject matter experts

    🔄 iteration 🤖 development support 🛠 refactoring 🧠 mindset 🏢 in-house development
  27. github.com/explosion/spacy-layout text-based contents layout sections section type document layout bounding

    box content, tokens, o sets process and create a spaCy Doc spaCy + Docling
  28. github.com/explosion/spacy-layout text-based contents layout sections section type document layout bounding

    box content, tokens, o sets process and create a spaCy Doc spaCy + Docling annotation in context
  29. At their core, many NLP systems consist of flat classifications.

    You can shove them into a single prompt, or you can decompose them into smaller pieces. Many classification tasks are straightforward to solve nowadays – but they become vastly more complicated if one model needs to do them all at once. explosion.ai/blog/human-in-the-loop-distillation
  30. Reason and refactor. The key to success lies in your

    data and may surprise you! Think beyond chat bots. You don’t want to build a “window-knocking machine”.
  31. Reason and refactor. The key to success lies in your

    data and may surprise you! LLM Stay ambitious. Don’t compromise on best practices, e iciency and privacy. Think beyond chat bots. You don’t want to build a “window-knocking machine”.