AI EXPO & DEMO HALL

In-Person | Keynote | Machine Learning | All Levels

Large Language Models (LLMs) like ChatGPT have taken the world by storm with their ability to answer questions, write essays and even compose lyrics. These tools have profound implications for industries like financial services, retail, and healthcare. However, most organizations have yet to take advantage of LLMs. Why? The amount of compute, data, and knowledge required to build a proprietary model is daunting for many, yet the alternative is reliance on LLMs that are only accessible behind API paywalls and compromising your data privacy. In this keynote, we will explore why training, deploying, and owning your own LLM is critical (and at times even imperative). We will discuss how you can train and deploy your own models, while protecting your data and your business IP. Spoiler alert: in contrast to common wisdom, ownership of your own LLM is within reach for most organizations, and provides major benefits in increased security, flexibility, and accuracy…more details

Virtual | Keynote | Machine Learning | All Levels

AI is red hot, but in practice many projects still fail. This talk will cover some of the key things you need to know to succeed, including:
– What current AI is and is not good for
– The difference between a demo and a product
– Pitfalls to avoid
– Organizing AI teams…more details

Virtual | Keynote | Machine Learning | All Tracks | All Levels

Modern medicine has given us effective tools to treat some of the most significant and burdensome diseases. At the same time, it is becoming consistently more challenging and more expensive to develop new therapeutics. A key factor in this trend is that we simply don’t understand the underlying biology of disease, and which interventions might meaningfully modulate clinical outcomes and in which patients. To achieve this goal, we are bringing together large amounts of high content data, taken both from humans and from human-derived cellular systems generated in our own lab…more details

Demo Talk | In-person

Thinking about incorporating relationships into your data to improve predictions and machine
learning models? Maybe you are creating a knowledge graph or looking for a way to improve
customer 360, fraud detection, or supply chain performance. Relationships are highly predictive
of behavior. With graphs, they’re embedded in the data itself, making it easy to unlock and add
predictive capabilities to your existing practices.
Join us for a demo to learn why graph databases are a top choice for scalable analytics, intelligent
app development, and advanced AI/ML pipelines. We’ll showcase graphs using Neo4j’s enterprise-
ready graph data platform. You’ll see firsthand how easy it is to get started and we’ll highlight a
graph use case using Neo4j’s cloud platform for Graph Data Science. All attendees will get a link to
download and try Neo4j for free using your own data.”

Demo Talk | In-person

Oracle Cloud Infrastructure (OCI) is proud to showcase a new product demo of Stable Diffusion for game content creation using popular user interfaces and the 3D modeling tool Blender. OCI’s demo of Stable Diffusion is powered by NVIDIA A10 Tensor Core GPUs in the Oracle cloud. Stable Diffusion, an innovative deep learning model released in 2022, has been primarily used for generating detailed images based on text descriptions. However, its capabilities extend to creating game textures, models, depth maps, skins, and other game content.

Diffusion models can even be utilized for other modalities, enabling tasks as diverse as music generation. The combination of Stable Diffusion and Blender allows artists to create high-quality game assets with complete control over the creative process while benefitting from quicker creative iterations. Artists can further train Stable Diffusion on their individual styles and develop complete workflows that allow for greater creative freedom and flexibility in game development.

Demo Talk | In-person

In the Python open-source ecosystem, many packages are available that cater to:
– the building of great algorithms
– the visualization of data

Despite this, over 85% of Data Science Pilots remain pilots and do not make it to the production stage.

With Taipy, a new open-source Python framework, Data Scientists/Python Developers are able to build great pilots as well as stunning production-ready applications for end-users.
Taipy provides two independent modules: Taipy GUI and Taipy Core.

In this talk, we will demonstrate how:
Taipy-GUI goes way beyond the capabilities of the standard graphical stack: Gradio, Streamlit, Dash, etc.
Taipy Core fills a void in the standard Python back-end stack.

Demo Talk | In-person

Integrating and unifying data from diverse sources is foundational to AI and ML workflows. This workshop will demonstrate how Anzo’s knowledge graph platform can create an enterprise-scale knowledge graph from several sources – setting organizations up for sustainable success with collective intelligence. During this workshop, users will:

• Create a sample knowledge graph from several sources
• Demonstrate flexible data preparation for training datasets
• Analyze the knowledge graph with native visualizations and graph algorithms
• Connect to the knowledge graph for additional data science operations

From its hyper agile in-memory MPP graph engine to its point-and-click user experience and open flexible architecture, Anzo transcends the limitations of traditional knowledge graphs and gives you all the capabilities and flexibilities that complex, enterprise-scale solutions need.

Join this demo to see why Anzo might be the solution you need.

Demo Talk | Virtual

Modeling time series data is difficult due to its large quantities and constantly evolving nature. Existing techniques have limitations in scalability, agility, explainability, and accuracy. Despite 50 years of research, current techniques often fall short when applied to time series data. The Tangent Information Modeler (TIM) offers a game-changing approach with efficient and effective feature engineering based on Information Geometry. This multivariate modeling co-pilot can handle a wider range of time series use cases with award-winning results and incredible performance.

During this demo session we will showcase how best-in-class and very transparent time series models can be built with just one iteration through the data. We will cover several concrete use cases for advanced time series forecasting, anomaly detection and root cause analysis.

Demo Talk | Virtual

Experienced machine learning engineers and data scientists care about ways to easily get their models up and running quickly and share ML assets across teams for collaboration. Collaborate and streamline the management of thousands of models across teams with new, innovative features in Azure Machine Learning. Come and join us in this interactive session with our product experts and get your questions answered on the latest capabilities in Azure Machine Learning!

Demo Talk | In-person

Great AI starts with great features. While the modern data stack has made self-service ingestion and consumption a reality for BI, AI data remains a huge challenge. Feature engineering is non-standard, ML pipelines are manual, and data governance is a nightmare – limiting the scalability you can achieve with AI. We will discuss the shortcomings of the modern data stack for AI, and practical approaches for creating a self-service data environment for data scientists. Learn about strategies to accelerate feature engineering and experimentation, shorten the time to deploy feature pipelines, and govern the data and infrastructure. And ultimately, to truly scale AI in your organization.

Presented by Razi Raziuddin, Youssef Idelcaid, FeatureByte

Demo Talk | In-person

Have you ever wandered into a department store and wondered if the item you are looking at is on sale? What if you had the ability to scan the item ‘yourself’ and determine if it was discounted? That ‘ability’ would make your shopping experience easier as you would not have to wait for customer service to look up your item. You would have the ability to find merchandise discounts, yourself, as you browse through the department store!

In this session, we will examine the model development, training, and deployment process we used to create a discount coupon app that is executed from an edge device – your phone. Join us as we walk you through the model development and deployment lifecycle, including topics of object vision, containerizing models, data streaming, and MLOps.

At the end of the session, you can try the app yourself.

Demo Talk | Virtual

With DALLE and ChatGPT, we have reached incredible capabilities and results, fundamentally changing our ability to tap into and leverage unstructured data in machine learning. With that said, general architectural understanding and intuition into how these models make decisions do not translate into minute detail interpretability.

We’re at a crossroads. This new “”””breed”””” of ML applications is here to stay, and unstructured data is only growing, but they are black boxes, and black boxes fail silently. So how can we as practitioners leverage NLP and vision while enjoying similar monitoring, interpretability, and explainability available to their tabular counterparts?

In this talk, we will introduce Elemeta, our OSS meta-feature extractor library in Python, which applies a structured approach to unstructured data by extracting information from text and images to create enriched tabular representations. With Elemeta, practitioners can utilize structured ML monitoring techniques in addition to the typical latent embedding visualizations and engineer alternative features to be utilized in simpler models such as decision trees.

In this talk, we’ll introduce you to Elemeta through a live notebook example and explain how it can be applied to text and images.

Demo Talk | In-person

Modeling time series data is difficult due to its large quantities and constantly evolving nature. Existing techniques have limitations in scalability, agility, explainability, and accuracy. Despite 50 years of research, current techniques often fall short when applied to time series data. The Tangent Information Modeler (TIM) offers a game-changing approach with efficient and effective feature engineering based on Information Geometry. This multivariate modeling co-pilot can handle a wider range of time series use cases with award-winning results and incredible performance.

During this demo session we will showcase how best-in-class and very transparent time series models can be built with just one iteration through the data. We will cover several concrete use cases for advanced time series forecasting, anomaly detection and root cause analysis.

Demo Talk | In-person

Yet another meeting to discuss nuisance hairsplitting details for your data taxonomies and keywords list?

It shouldn’t take a team of domain experts, Excel specialists, Python developers, and Data Scientists weeks or months to build it. It is a simple problem that requires a simple solution.

You should be able to quickly and accurately analyze contracts, customer comments, and any other text-based content while easily building explainable NLP models.

Stop scrubbing through volumes of data to find key examples and then reducing the content to specific keywords and variations.

Join us as we explore a new and exciting solution using human language to easily develop ontologies, data taxonomies, and keyword lists, which you can share across your business with just a few simple clicks.

Accelerate these NLP tasks in every project and help eliminate those long-drawn meetings to discuss keywords for data taxonomies. Unless you enjoy those nuisance meetings 🙂

Demo Talk | In-person

The process of investigating model performance issues in production environments has long been plagued by complexity, inefficiency, and limited success. In this demo talk, we unveil a new paradigm for streamlining performance diagnostics and performing effective RCA. Our Root Cause Analysis enables you to effortlessly slice and dice production data and identify previously hidden relationships and valuable insights from your raw data. Easily collaborate to find when, where, and why issues originated to expedite response and remediation. Investigate any use case, any production issue, in every model type, and turn deep insights into actionable success.

Demo Talk | Virtual

Integrating and unifying data from diverse sources is foundational to AI and ML workflows. This workshop will demonstrate how Anzo’s knowledge graph platform can create an enterprise-scale knowledge graph from several sources – setting organizations up for sustainable success with collective intelligence. During this workshop, users will:

• Create a sample knowledge graph from several sources
• Demonstrate flexible data preparation for training datasets
• Analyze the knowledge graph with native visualizations and graph algorithms
• Connect to the knowledge graph for additional data science operations

From its hyper agile in-memory MPP graph engine to its point-and-click user experience and open flexible architecture, Anzo transcends the limitations of traditional knowledge graphs and gives you all the capabilities and flexibilities that complex, enterprise-scale solutions need.

Join this demo to see why Anzo might be the solution you need.

Demo Talk | In-person

Learn why the truly open-source HPCC Systems platform is better at Big Data and offers an end-to-end solution for Developers and Data Scientists. Learn how ECL can empower you to build powerful data queries with ease. HPCC Systems, a comprehensive and dedicated data lake platform makes combining different types of data easier and faster than competing platforms — even data stored in massive, mixed schema data lakes — and it scales very quickly as your data needs grow. Topics include HPCC Architecture, Embedded Languages and external data stores, Machine Learning Library, Visualization, Application Security and more.

Demo Talk | In-person

We will be using the plotly library to create visualizations in S&P Global Marketplace Workbench, which is powered by Databricks, and showcasing a Databricks Dashboard from the different charts. I would say this demo talk is best suited for beginner – intermediate audience.

Demo Talk | In-person (AI Startup Showcase)

As companies collect larger and larger amounts of data and apply more complex ML models, the time and resources required to build and maintain the models continues to grow. In fact, in the past decade, training compute time has been growing at a staggering pace of 10x per year! But do we really need so much data to build better models?

In this talk, we will walk you through DataHeroes’ Python-based framework that uses a unique data reduction methodology to reduce your dataset size by orders of magnitude while maintaining the statistical properties and corner cases of the full dataset. We will demonstrate how having the reduced dataset makes it significantly easier and faster to clean the data and train and tune the model, which produces a better and more accurate model, at a fraction of the time and cost.

Demo Talk | In-person

When putting models into production it’s critical to know how they’re performing over time. As the last mile of the data pipeline, models can be impacted by a variety of issues, often outside the control of the data science team. “Observability” promises to help teams detect and prevent issues that could impact their models—but what is observability vs. data observability vs. ML observability? Get practical answers and recommendations from Kyle Kirwan, former product leader for Uber’s metadata tools, and founder of data observability company, Bigeye.

Demo Talk | In-person

In this 25-minute demo, we will explore the top 5 cool tricks of Delta for data scientists and discuss why your data lake should be a Delta Lake. Delta Lake is an open-source storage layer that brings reliability to data lakes by providing ACID transactions, scalable metadata handling, and data versioning. We will first introduce the concept of Delta Lake and explain how it helps data scientists to manage their data pipelines with ease. We will then dive into the top 5 cool tricks of Delta Lake, which include performance optimizations, time travel, schema enforcement, automatic data merging, and data validation. We will demonstrate these tricks using real-world examples and show how they can simplify your data pipeline and reduce your development time. By the end of this talk, you will have a better understanding of Delta Lake’s features and how it can help you to manage your data lake efficiently. You will also have learned about the benefits of using Delta Lake and why it’s a must-have for data scientists working with large data sets.

Demo Talk | In-person

Location intelligence can provide valuable insights by leveraging geospatial data in machine
learning. This demonstration will showcase how machine learning and location information can
work together to help organizations extract more value from their data. We will use a
comprehensive suite of geocoding, spatial analytics, and data enrichment capabilities to
visualize and analyze data, identify patterns, and derive insights that can be used to make
informed business decisions.
In this session, we will use Amazon SageMaker to train a machine-learning model using
property attributes, historical weather data, and fire data. The goal will be to predict the fire
risk for a property. You will see how quickly and efficiently we can build and train a machine-
learning model using various algorithms, such as decision trees and neural networks, to find the
best approach for our dataset.
While this demo example will highlight fire predictions, these location intelligence solutions can
be applied across multiple industries, including financial services, telecommunications,
insurance, retail, real estate, and more. We look forward to discussing how you can leverage
geospatial data in your machine-learning models.

Demo Talk | Virtual

Proper tracking is crucial for ensuring the reproducibility of results obtained during model development and fostering effective collaboration among multiple developers on a machine learning project. In this talk, Kristen will discuss the process of developing a dog detection system using YOLOv8 on-edge devices and the role of Comet, an experiment management platform, in handling the intricacies of the project.

Kristen will guide you through the entire process, from generating a data artifact to deploying the model, emphasizing the benefits of utilizing Comet at each stage. She will showcase how Comet was employed to monitor experiment metrics, visualize model performance, and illustrate the ease with which the selected model can be tracked in production. Participants will gain valuable insights on how to leverage an experiment tracking and monitoring solution like Comet to enhance their model development process, making it more transparent and reproducible.

Demo Talk | In-person

MLRun is an open-source MLOps orchestration framework. It exists to accelerate the integration of AI/ML applications into existing business workflows. MLRun introduces Data Scientists to a simple Python SDK that transforms their code into a production-quality application. It does so by abstracting the many layers involved in the MLOps pipeline. Developers can build, test, and tune their work anywhere and leverage MLRun to integrate with other components of their business workflow.

The capabilities of MLRun are extensive, and we will cover the basics to get you started. You will leave this session with enough information to:

• Get started with MLRun, on your own, in 10 minutes, so you can automate and accelerate your path to production and have your first AI app running in 20 minutes
• Run local move to Kubernetes
• Understand how your Python code can run as a Kubernetes job with no code changes
• Track your experiments
• Get an introduction to advanced MLOps topics using MLRun”

Demo Talk | Virtual

This technical talk delves into the paradigm shift from model-centric to data-centric AI, emphasizing the importance of data quality in improving machine learning outcomes. We will explore the current AI landscape and discuss the reasons behind this shift. Focusing on the Pachyderm platform for data-driven processing and versioning, attendees will learn practical steps and principles to streamline their data-centric AI efforts. This talk aims to equip practitioners with the knowledge and tools necessary to harness AI’s full potential by embracing a data-driven approach and leveraging Pachyderm’s innovative platform.

Demo Talk | In-person

With DALLE and ChatGPT, we have reached incredible capabilities and results, fundamentally changing our ability to tap into and leverage unstructured data in machine learning. With that said, general architectural understanding and intuition into how these models make decisions do not translate into minute detail interpretability.

We’re at a crossroads. This new “”””breed”””” of ML applications is here to stay, and unstructured data is only growing, but they are black boxes, and black boxes fail silently. So how can we as practitioners leverage NLP and vision while enjoying similar monitoring, interpretability, and explainability available to their tabular counterparts?

In this talk, we will introduce Elemeta, our OSS meta-feature extractor library in Python, which applies a structured approach to unstructured data by extracting information from text and images to create enriched tabular representations. With Elemeta, practitioners can utilize structured ML monitoring techniques in addition to the typical latent embedding visualizations and engineer alternative features to be utilized in simpler models such as decision trees.

In this talk, we’ll introduce you to Elemeta through a live notebook example and explain how it can be applied to text and images.

Demo Talk | In-person

Pressure to maximize ROI and reduce costs continues to grow for ML organizations. Data labeling is often one of the line items under the most scrutiny, so how do you get more efficient with your spending?

In this demo, we’ll walk through several ways that you can need fewer labels and make smarter choices on what you label through a combination of automation and human-in-the-loop.

Our clients using these techniques were able to:

• Reduce the amount of data labeled by up to 90%
• Find frames with rare objects within huge datasets
• Optimize annotation guidelines to improve their model performance

Demo Talk | In-person

Proper tracking is crucial for ensuring the reproducibility of results obtained during model development and fostering effective collaboration among multiple developers on a machine learning project. In this talk, Kristen will discuss the process of developing a dog detection system using YOLOv8 on-edge devices and the role of Comet, an experiment management platform, in handling the intricacies of the project.

Kristen will guide you through the entire process, from generating a data artifact to deploying the model, emphasizing the benefits of utilizing Comet at each stage. She will showcase how Comet was employed to monitor experiment metrics, visualize model performance, and illustrate the ease with which the selected model can be tracked in production. Participants will gain valuable insights on how to leverage an experiment tracking and monitoring solution like Comet to enhance their model development process, making it more transparent and reproducible.

Demo Talk | In-person

When OpenAI launched ChatGPT at the end of 2022, more than one million people had tried the model in just a week and this trend has only accelerated with ChatGPT recently reaching 100 Million monthly users. It’s clear that NLP and Generative Large Language Models are becoming mainstream. In this talk, you will learn how to enable ChatGPT on your own data with vector search.

Generative LLMs like ChatGPT are trained on huge datasets of open data from the internet. This enables them to have vast amounts of general knowledge about the world and natural language. However, there is one disadvantage: once trained, you can only use Generative Pretrained Transformers (GPT) only on the data that it was trained on. When you ask what today’s news is, ChatGPT can’t answer that question.

In order to benefit from the capabilities of LLMs like ChatGPT in real-life use cases, it would be ideal if we could apply the generative power to new or custom data, such as a chatbot for your e-commerce platform which has knowledge about the products you sell, your detailed return policies or specific promotions currently going on. This becomes possible if we combine ChatGPT with a vector search engine! Integrating a generative LLM with a vector search engine allows you to filter through your entire personal database and search for information relevant to a prompt, which can then be provided to ChatGPT along with the prompt. This framework allows you to harness the LLM model’s power to solve tasks grounded in the context of your own data!

In this demo, we will show you how ChatGPT can be implemented with the open-source vector search engine Weaviate in live demos.

You will leave this talk with a solid understanding of how you can enable ChatGPT on your own data using vector search. Whether you’re a data scientist, developer, or NLP enthusiast, this talk will provide valuable insights and practical skills for enhancing your NLP projects with vector search and OpenAI’s ChatGPT.

Demo Talk | In-person

This technical talk delves into the paradigm shift from model-centric to data-centric AI, emphasizing the importance of data quality in improving machine learning outcomes. We will explore the current AI landscape and discuss the reasons behind this shift. Focusing on the Pachyderm platform for data-driven processing and versioning, attendees will learn practical steps and principles to streamline their data-centric AI efforts. This talk aims to equip practitioners with the knowledge and tools necessary to harness AI’s full potential by embracing a data-driven approach and leveraging Pachyderm’s innovative platform.

Demo Talk | In-person

Beacon Analytics helps customers transition from rigid and monolithic data solutions to flexible microservices architecture, enabling better performance and faster access to critical information. By breaking up data into smaller, independent services, customers gain greater access and modification capabilities. The team recommends using the Polars library, which is based on Apache Arrow, in combination with Dash Plotly to create easy to maintain, high-performance solutions at an excellent price-to-performance ratio. Join Danny Bharat, Senior Vice President of Analytics at Cedric Millar and co-founder of Beacon Analytics, as he shares how his team’s innovative approach to data solutions allows them to build comprehensive 360° intelligence and deliver actionable insights. Beacon Analytics empowers customers to achieve success in a rapidly changing business and technology landscape by utilizing schema-on-read approaches, unstructured data storage, and on-the-fly analysis and transformation.