7:00 午後 - 8:00 午後 (UTC)

In this session led by Steve Notley from the Field Engineering team at Wallaroo.AI you will learn how to deploy, serve, optimize and observe Computer Vision models in production both in the cloud and at the Edge.

9:00 午後 - 10:00 午後 (UTC)

Learn how to get instant data insights and drift detection alerts from pipelines deployed at edge locations without any operational overhead. See how you to create aggregated drift detection assays on inputs and outputs for the pipelines deployed in an ML operations center & all the edge deployments Why should I attend? You will leave this session with a comprehensive understanding about how to observe ML models deployed at the edge for data drift and take corrective action for under performing models. Try this Computer Vision model and other common AI use cases using the Wallaroo.AI Azure Inference Server Freemium Offer on Azure Marketplace (https://portal.azure.com) and also try the Free Wallaroo.AI Community Edition (https://portal.wallaroo.community/)

9:00 午後 - 10:00 午後 (UTC)

About this session: Learn how to automate and scale ML workflows with workload orchestration for both batch and real-time inference serving. In this session we will dive into the steps to deploy, automate and scale recurring production ML workloads that can ingest data from predefined data sources to run inferences, chain pipelines, and send inference results to predefined destinations to analyze model insights and assess business outcomes. Why should I attend? You will leave this session with a comprehensive understanding about how to manage and orchestrate model automation for batch and real-time inference serving scenarios Resources: Try this Computer Vision model and other common AI use cases using the Wallaroo.AI Azure Inference Server Freemium Offer on Azure Marketplace (https://portal.azure.com) and also try the Free Wallaroo.AI Community Edition (https://portal.wallaroo.community/)

6:00 午後 - 7:00 午後 (UTC)

Getting AI models into production is hard. To begin with there are many different model building frameworks on the market to choose from and each comes with a unique way to package them for production adding to the complexity for getting to the production start line. In this session we will cover getting the model framework of your choice to production in a standardized way as well as using ML model registry for version control as the models move between training, production, monitoring, and deployment.

4:00 午後 - 5:00 午後 (UTC)

Congratulations! You have deployed your AI models to the Edge. How do you make sure they are performing the way they were intended to? If they are not, what can you do about it? In this session we will deep dive into capturing observability data on Edge deployments even when the network connection is intermittent or has limited bandwidth and have the ability to return the data during specific time periods to run observability on this data. We will show how Data scientists in the Model Operations Center can monitor drift for models deployed for a specific edge location or a group of edge locations and take action on underperforming models by hot swapping in better performing models. Try this Computer Vision model and other common AI use cases using the Wallaroo.AI Azure Inference Server Freemium Offer on Azure Marketplace (https://aka.ms/Wallaroo-Inference) and also try the Free Wallaroo.AI Community Edition (https://aka.ms/Wallaroo.AI-Free)

6:00 午後 - 7:00 午後 (UTC)

In this session we will dive into deploying LLMs to Production Inference Endpoints and then putting in place automated monitoring metrics and alerts to help track model performance and suppress potential output issues such as toxicity. We will also cover the process of optimizing LLMs using RAG, for relevant, accurate, and useful outputs. You will leave this session with a comprehensive understanding about deploying LLMs to production and monitoring the models for issues such as Toxicity, relevance, and accuracy. Try this other common AI use cases using the Wallaroo.AI Azure Inference Server Freemium Offer on Azure Marketplace and also try the Free Wallaroo.AI Community Edition

6:00 午後 - 7:00 午後 (UTC)

In this session we will dive into the details for how to build, deploy, and optimize custom Large Language Models (LLMs) for production inference environments This session will cover the key steps for Custom LLMs (LLama), focusing on: Why custom LLM's? Inference Performance Optimization Harmful language Detection