List of videos

Full title: Building LLM-based Agents: How to develop smart NLP-driven apps with Haystack Sponsor: deepset Large Language Models (LLMs), like ChatGPT, have taken the Internet by storm. People are amazed at how these models can generate language with the snapshot of general world knowledge they have accumulated. However, in many real-world use cases a snapshot of general world knowledge is not enough. How do you equip them with the relevant tools to solve a task - like SQL, a CRM or web search? Or for a task, how do you control which tool is used or which knowledge base is referred to? Using LLMs as so-called “Agents” allows you to use them as a decision maker in your application that react to all sorts of user requests. Given a user request, the agents can create an action plan. In this talk, we will learn how to build agent-driven applications with Haystack. We will show how to build an agent and connect it to other tools or knowledge bases. We will illustrate the concept with practical use cases and examples. Each step will be accompanied by code examples. By the end of the talk, you will have seen these concepts applied in practice, and you will be able to build an agent-driven application for your own use case.

Sponsor: JetBrains Django and Python make fullstack and API web projects a breeze. But as Python has matured, significant tooling has risen to improve the development experience (DX). Can you use this tooling, in a modern editor and IDE, to stay in the flow and make your development…joyful? In this session we’ll put PyCharm to work, at helping us work. Navigation, refactoring, autocomplete – the usual suspects. We’ll also see “test-first” development to stay in the IDE, plus how this can apply to frontends. Finally, we’ll follow along with topics from Adam Johnson’s book “Boost Your Django DX”, with a surprise at the end.

By Guillaume Dequenne Sponsor: Sonar Dealing with legacy codebases can be a chore. In this workshop, we talk about modernizing an old Flask application one pull request at a time while incorporating Python best practices. We will talk about how to integrate Code Quality tools in your workflow and in your IDE, so that every pull request is checked to good standards. By the end of this Workshop, we hope everyone can set up a workflow that removes technical debt over time and makes their codebases sustainable. This presentation can be followed interactively. In order to do so, make sure that you meet the following prerequisites: - Have an account on GitHub and be logged-in - Fork the workshop GitHub repository in your personal account, with all its branches (by default, only the main branch is forked): https://github.com/SonarSource/pycon-sonar-workshop - Optionally, have Git and an IDE (PyCharm/VSCode) to clone the application locally

Presented by: - Pradyun Gedam - Bernat Gabor - Laszlo Kiss Kollar - Mario Corchero - Matt Wozniski - Pablo Galindo Salgado Sponsor: Bloomberg Join this talk where we will briefly introduce Bloomberg and have some of our engineers discuss their engagement in the Python Open Source ecosystem. We will also present some exciting troubleshooting tools that are widely used at Bloomberg that we are publishing as open source. You will leave this talk having learned about the technical details and new features related to these open source tools, which you might use daily in the future!

Full title: Breaking Boundaries: Advancements in High-Performance AI/ML through PyTorch's Python Compiler By Sam Gross, Justin Jeffress and Suraj Subramanian Sponsor: Meta As GPUs continue to become faster, PyTorch, one of the most widely used frameworks in AI/ML has faced challenges keeping up with performance demands. To mitigate this, parts of PyTorch have been moved into C++. This approach goes against the original intent of PyTorch as a Python-based framework and complicates contributions from the community. The PyTorch development team recognized the need to address these challenges while maintaining PyTorch's Python roots and set ambitious goals to improve performance, decrease memory usage, enable state-of-the-art distributed capabilities, and ensure more PyTorch code is written in Python. To achieve these goals, they developed a Python compiler. Attendees of this talk will get an inside look at how the PyTorch development team approached these challenges and implemented their innovative solution to achieve a 43% speedup in performance. We will discuss the benefits and challenges of this approach, as well as the techniques and technologies used to build the PyTorch Python compiler. This talk will provide valuable insights into the development process of and offer attendees a deeper understanding of how PyTorch continues to evolve and innovate.

By Shana Matthews and Indragie Karunaratne Sponsor: Sentry Most Python devs are familiar with built-in profiling tools like cProfile, but the world of profilers has expanded rapidly. In this hands-on workshop we'll explore various open-source profiling technologies with different overhead and accuracy tradeoffs, as well as several ways of visualizing profile data like flow charts, call trees, and flamegraphs.

Full title: Python Meets Heterogeneous Computing: An Exploration of Distributed Computations By William Cunningham and Santosh Kumar Radha Sponsor: Covalent In this talk, we will delve into the exciting world of advanced hardware and its use for distributed computations. As the heterogeneous computing landscape evolves with the introduction of quantum computers, GPUs, and specialized hardware, we will explore the interesting patterns in Python that allow us to interact with this heterogeneity and maximize performance effectively. The talk will address the challenges of distributed computations in the heterogeneous computing era, including monitoring real-time calculations, rapid iteration, prototyping in complex experiments, and ensuring smooth production runs with long access queues in specialized hardware. However, we will not only highlight the difficulties but also share valuable strategies for overcoming these obstacles and achieving optimal performance in these environments using open-source tools.

By Cody De Arkland Sponsor: LaunchDarkly Everyone acts like the party is in the Frontend, but the Backend is what keeps your platforms running. Python has a number of options for building backend API's but the most common are Flask and Django. API's are the backbone of most applications, and shipping new ones is often risky - but they don't have to be. In this session, Cody will take you through where he started his coding journey with Python and Flask, go hands on with examples of backend API's, and show how you can ship faster and safer using feature flags.

By Jason Koo and Alison Cossette Sponsor: neo4j Become a Rebel and learn how to create a hyperspace navigation app to avoid Imperial patrols using the power of graphs in Python! Say goodbye to complex if-then statements and embrace a more elegant and flexible approach. In this talk we'll introduce graphs as a data structure, how to model this kind of data, and how to use them in place of more complicated logic code. Discover how graphs can simplify decision-making logic in your code for improved readability and extensibility. Join us on a journey to break free from the empire of if-then statements and unlock the full potential of your data. Expand your toolset and liberate your code!