Part 1a. What is ai / ml?

Part 1b. Limitations of current approach

The current way of doing ML is necessary but limited. If we truly want "intelligence", we need a new way of doing things i.e. an AI that can think beyond its training, an AI that requires much less data, and an AI that is much more adaptable than current AI. Here we give case studies of current AI systems and biology highlighting the limitations and potential of what a real intelligent system looks like.

Part 1c. The upcoming AI / ML winter

Part 2a. What does python have to do  with ML?

Part 2b. The new approach to ML

What use is Python at all? Is it possible to conceptualize ML with or without python? In this section, I explain how Python is used to conceive AI/ML. Then I propose a new way of approaching the problem. I explain the various ways Python is used in ML from deep learning, reinforcement learning etc.

Part 3. A new way of thinking

In this section, I propose a different way of thinking of the problem and propose a solution.

NuMat Technologies is a team of chemists, chemical engineers, and computer scientists developing advanced materials to remove toxic chemicals and greenhouse gases from air, water, and more.

Founded by computer scientists and working with a new material class that lends itself well to computational design, NuMat puts computation at the forefront of the business. Whether this is developing automated material design "recommendation engines", building robotics for high-throughput experimentation, or maintaining our in-house enterprise resource planning applications, NuMat's computational team touches every aspect of the company.

Expect to see async and advanced communication protocols in robotics, Django+SQL with our ERP applications, and HPC management software like Dask and Jupyter in our computational material design.

Narrative Science is a data storytelling company that has been dynamically writing stories and reports for over a decade

This talk will go over how we are using Python and its rich ecosystem to move towards a microservices architecture that will create a more scalable and fault tolerant product.

Zoro is an online distributor of products for B2B customers, focused on helping small businesses easily find what they need to grow and maintain their businesses. Today, we have over eight million products available—and that number is expected to keep growing. We work with third-party suppliers to provide products and fulfill orders for our customers.

Zoro uses Python with Django for its ecomerce site, as well as for data science, ETLs, and microservices.

At JFrog, we are making endless software versions a thing of the past, with liquid software that flows continuously and automatically from build all the way to deployment.  With this in mind, we’ve developed the world’s first universal artifact management platform, ushering in a new era in DevOps – Continuous Updates. Ten years later, with thousands of customers, and millions of users globally, JFrog has become the “Database of DevOps” and de-facto standard in release and update management.

JFrog embraces the Python language for multiple uses cases and technology solutions including provisioning machines, tooling for Pipelines, creating machine learning models, securing Python modules, and even Python-based micro services in the JFrog Platform.

Nielsen is a global leader in audience insights, data and analytics, shaping the future of media. Nielsen uses Python to bridge the gap between model development, validation and deployment into production data pipelines to accelerate creation and evolution of analytics products.

Easy to build Python Dashboards using Financial data APIs

Automated tests are a great way to iterate fast and ensure features didn't break. This talk discusses how to speed up your builds and dev cycle even more by running tests asynchronously using a pytest plugin called asyncio-cooperative.

Building machine learning (ML) models is faster and easier now than ever before. The proliferation of open-source libraries means data scientists can leverage cutting-edge pre-trained models in just a few lines of code. Yet it remains true that most ML models never make it to production. Why? Because making it to production (and staying in production) are about more than just model and code quality. In particular, this talk will discuss how MLOps can greatly accelerate and increase the chances of model success.

Specifically, the talk will walk through the full ML lifecycle and answer: What is MLOps? Why is it important? How can MLOps infrastructure be set up quickly, easily, and with open source tools? How can the system be designed in a user-friendly way, but without too much magic? How can user adoption be accelerated?

While its expected that data-science-related professionals will garner the most value from this talk, no prior MLOps/ML background is required to understand the contents of the talk.