Would you like to do a natural language processing project but feel overwhelmed by all the talk of ChatGPT, transformer models and text embeddings? Would you like to understand how you can take a set of raw texts and put them into a form that an AI model will understand?In this talk, you'll learn some of the theory behind two of the most widely used techniques in natural language processing today: word embeddings and large language models. You'll follow a practical demonstration of how you can use these techniques yourself, in which you'll see how to build a clickbait headline classifier in Python with user-friendly packages like `gensim` and `transformers`.By the end of this talk, you’ll have an understanding of why each technique works, and the advantages and disadvantages of using each of them. Even if you haven't done any machine learning before, you'll gain enough knowledge to go home and start experimenting with your own natural language processing project.

There has been a huge rise in the use of graphics processing units (GPUs) for general purpose computing outside of purely visual applications, be it for accelerating scientific research simulations, or training artificial intelligences in an extensive range of fields, from analysing medical images to predicting new proteins. This talk will look at some of the Python-based software, tools and frameworks that are available to help developers and researchers quickly utilise the full abilities of GPU acceleration technology in their work.

Let's demystify Gen AI and see how we can apply it to fun, approachable solutions, with a magical twist. We'll first explore how vector embeddings and LLMs work, before we set off to build our search solution (with a live demo).

Using the Elastic Python clients we first create indexes for Harry Potter characters, and film subtitles. We can import compatible 3rd party LLMs through an enriching pipeline; allowing us to add sentiment analysis and embeddings to our text. We’ll build a semantic search engine that can browse the books better than the ultimate fan.

The advent of real-time data processing has redefined how businesses operate, but the journey to real-time has its challenges. Many organizations find themselves juggling separate workflows and teams for batch and streaming data processing, each presenting unique difficulties. Batch processing, while effective for large datasets, can be costly, slow, and not well-suited for API integration. On the other hand, streaming, despite its speed and low latency, often has restricted functionality. This talk explores how Pathway is an open framework in Python for real-time data processing that provides a unified platform for batch and streaming. You will see with examples how simple it is to make your batch code run in streaming and learn how to build a data processing pipeline that ingests data from various data sources, processes, analyzes, and sends it to output streams in real-time.