Accelerate App Development with Intel’s AI PC Development Kit
Learn how to jumpstart the development of AI PC applications for your customers by working through hands on exercises using the AI PC Development Kit. This lab steps you through using key software tools for performing AI Inference on an AI PC with an Intel® Core Ultra processor. The demonstration will include OpenVINO™ toolkit, ONNX with OpenVINO™ Execution Provider, and WebNN.
Accelerate Compression with Intel® Quick Assist Technology
Rapidly expanding data footprints increase computational costs for both CSPs and enterprise customers. Compression is a common approach to reduce data footprint, but the associated CPU overhead and workload latency impact are high. This often pressures CSPs and enterprise customers to disable compression in key workloads. In this talk, we introduce hardware compression offload through Intel® Quick Assist Technology across Microsoft enterprise and cloud use cases in networking, storage, and SQL Server query engine. Detailed data analysis is shared to demonstrate performance improvements along with CPU cycle and data footprint reductions from the hardware offload.
AI Adopting Research Data Management using oneAPI & openVINO
The adoption of AI and ML in research is driving exponential growth in research data. Robust research data management (RDM) practices are crucial for ensuring regulatory compliance when building trustworthy AI systems. We discuss integrating containerization with oneAPI and openVINO AI frameworks, enabling UXL, and a cohesive RDM approach across advanced computing pipelines. A case study demonstrates enhancing data security, integrity, and reproducibility on CPUs/GPUs through this innovative strategy. This proposal furthers the initiative we initially presented at the oneAPI Devsummit for AI and HPC 2023 and will include relevant key points of the national research strategy 2025-2030.
AI for Daily Efficiency: The Productivity Revolution by kAI
Discover how kAI is revolutionizing productivity through innovative processes powered by Intel's cutting-edge infrastructure. Our advanced Large Language Model (LLM) enhances daily tasks and operational efficiency by leveraging Intel's CPUs and GPUs for fine-tuning and deployment. We will showcase scalable GenAI model deployment and efficient fine-tuning for specific use cases. Additionally, we'll discuss implementing automated pipelines and performance monitoring within cloud-native environments, all made possible through Intel's robust hardware solutions. Join us to explore how Intel's technology is integral to our success and the future of AI-driven productivity enhancements.
AI Media Enhancement without the Headache
With the exponential growth of video, the media and entertainment industry needs AI-based solutions to enhance user experience and reduce distribution costs. However, this can be challenging without a team of experts in both video and AI. In this session, we’ll demonstrate how AI-based media enhancements can be easily deployed using cloud-based offerings from AWS, or at the edge using the Converged Edge Media Platform (CEMP) and OpenShift from Red Hat. We’ll provide an overview of the key Intel technologies that enable these workloads, their applications in live broadcast, streaming, sports, and archives, and how to solve the challenges of a multi-workload, multi-platform edge environment.
AI/ML Driving Enhanced Safety & Security in Mixed-Criticality Automotive Systems
The rapid evolution of the automotive industry demands integration of ADAS, IVI, and Connected Vehicle Cloud systems within a robust, secure, and efficient computing framework. This presentation delves into the application of AI within real-time operating systems and hypervisors to manage mixed-criticality environments, ensuring peak safety and security in automotive systems. Utilizing AI-driven algorithms, we enhance predictive maintenance, anomaly detection, and real-time decision-making capabilities. Integration of DevOps practices facilitates seamless development, deployment, and servicing of embedded systems, promoting continuous improvement and rapid iteration. Our approach emphasizes the convergence of safety-critical and non-critical workloads on shared hardware platforms, optimizing resource utilization without compromising performance or reliability. AI methodologies extend to advanced sensor fusion techniques, combining data from cameras, LiDAR, and radar to create a comprehensive environmental model. This enables precise object detection, classification, and tracking essential for adaptive cruise control, lane-keeping assistance, and collision avoidance. AI-driven perception systems also improve performance in night vision and low-visibility scenarios, ensuring reliability across varying conditions. Attendees will gain insights into how AI methodologies bolster the resilience and efficiency of automotive edge devices. The presentation will explore architectural strategies and practical implementations, highlighting AI-enhanced safety protocols and security measures, driving advancements in autonomous driving technologies.
AI Myopia? GraphRAG Sees the Bigger Picture (vs. Vector RAG)
Is your AI just parroting data? Tired of narrow search results limiting its potential? Discover GraphRAG, the revolutionary approach that breaks free from Vector RAG's limitations. Unlike search engines, GraphRAG unlocks a universe of connected knowledge, giving your AI the big picture. See relationships, understand context, and unleash true knowledge retrieval for smarter, more insightful AI. Don't let your AI be myopic, join us and explore the power of GraphRAG!
Align LLMs with SeekrFlow: Finetune on Intel® Tiber™ Developer Cloud, Deploy on AI PC
In this workshop, we present a groundbreaking collaboration between Intel and Seekr, showcasing how SeekrFlow, a comprehensive platform for fine-tuning and deploying transformer-based models, can be used to align LLMs with domain-specific principles. By utilizing Intel® Tiber® Developer Cloud for fine-tuning and Intel’s upcoming client processor (code-named Lunar Lake) AI PC architecture for deployment, we offer an end-to-end solution for building and deploying principally aligned LLMs.
An AI stack: from compute infrastructure to LLM evaluation
With the release of ChatGPT, just over one year ago, large language models (LLMs) have taken the world by storm: they have enabled new applications, have exacerbated GPU shortage, and raised new questions about their answers' veracity. In this talk, I will present several projects I have been working on over the past three years, which are now part of an open-source stack for training, fine-tuning, serving, and evaluating LLMs. In this talk, I will focus on three projects: (i) Ray, a distributed framework for scaling ML workload, (ii) vLLM, a high-throughput inference engine for LLMs, and (iii) Chatbot Arena, a system to accurately benchmark LLMs.
Augmenting video with Camo Studio on Lunar Lake
Camo Studio is an award-winning app that helps anyone achieve incredible video for meetings, streams, and content creation, giving users precise controls and a range of AI-powered effects to perfect their image, with any camera and in any app. It was recently awarded a King’s Award for Innovation in the UK, featured in Satya Nadella’s keynote at Build 2024, and was an Apple Design Awards finalist for innovation in 2023. Taking advantage of Lunar Lake’s NPU is next on our roadmap to drive the app’s continuous ML workloads more efficiently without compromising performance or quality. The team is also working on the next major iteration of the app, Camo 3, and our first gen-AI features.
Azure Support for Intel® Trust Domain Extensions and the Open HCL Paravisor
This joint session with Microsoft will discuss Azure’s upcoming support for Intel® Trust Domain Extensions (Intel® TDX), including Public Preview and Intel® TDX support in Microsoft’s Open HCL open source paravisor project. This will include an overview of Intel® TDX and how Intel has worked with Azure to support TDX Partitioning using the Open HCL paravisor.
Beam Me Up, Intel: Exploring Lunar Lake and Arrow Lake Processors
Join us for an in-depth dive into Intel’s groundbreaking Lunar Lake and Arrow Lake processors. Discover why Lunar Lake is the cornerstone of the AI PC platform and represents a leap forward in processor design uniquely suited for AI applications. In this session, we’ll explore architectural details that enhance software development and learn how to effectively leverage silicon features to achieve unparalleled power efficiency and performance. Gain insights into our multi-engine approach, which not only delivers power-efficient performance but also scalability for AI client workloads.
Beyond LLMs: Foundation AI Models of Tomorrow
Discover how Intel Labs is leading the AI open ecosystem in solving problems in several domains. Large language models (LLMs) have transformed the public perception of AI. Intel Labs is already thinking about the next big frontier. We explore the scaling behavior of LLMs and how we can bring them to consumer grade hardware. We tackle open questions like how to better ground complex reasoning systems in knowledge and build disruptive solutions for areas like computational chemistry, chip design and video understanding. Learn more about these research vectors and how we are pushing the state of the art in AI to establish open-source tools and new open benchmarks for industry and academia.
BUD-E - Democratizing Education with Open Voice Assistants
In a world where quality education is often a privilege, BUD-E offers a hopeful glimpse into the future of learning. Developed through a collaboration between LAION e.V., the Max Planck Institute for Intelligent Systems Tübingen, and ELLIS, BUD-E is an innovative AI voice assistant designed to provide real-time, natural, and empathetic educational interactions. Imagine children in the US or Europe, as well as in remote areas of Africa, Pakistan or Afghanistan engaging with a personalized AI tutor that understands their needs and speaks their language. Learn more about our efforts can transform education, bridging gaps and providing high-quality tutoring for learners everywhere.
Building AI Apps with Web Neural Network (WebNN) on AI PC
This hands-on-lab invites developers to explore the emerging field of applications built with Web Neural Network (Web NN) on AI PCs. Participants will delve into the integration of models like speech-to-text, enabling seamless interaction with a Local Large Language Model (LLM). The session offers a hands-on approach to developing applications that harness the power of advanced AI technologies for the web and efficient user interactions. Additionally, this hands-on lab contributes to the requirements for the digital certificate of competency titled "Essentials for Building AI PC Applications," further validating your skills and expertise in this cutting-edge field.
Building a RAG Pipeline with Red Hat OPEA Validated Platform / RAG Appliance
Understand how to construct a Retrieval-Augmented Generation (RAG) pipeline using Red Hat's OPEA validated platforms / RAG Appliance (Gaudi). This session will cover the integration of data retrieval and generative models for enhanced AI solutions.
Building Local LLM Agents for Workflow Automation on AI PC
This hands-on lab equips developers with the skills to build local agents for research assistance, emphasizing practical expertise in Gen AI, NLP, and data management using external sources. The workshop starts with an introduction to local agents, progressing to technical construction aspects such as processing user requests and data presentation. It focuses on creating agents optimized for laptops with Intel® Core™ Ultra Processors, ensuring high performance. Participants will develop a local agent, applying iterative enhancements throughout the course. By the end of the lab, developers will be adept in local agent development, ready to build effective AI assistants for research and various other tasks. This hands-on lab also contributes to the requirements for the digital certificate of competency titled "Essentials for Building AI PC Applications," further validating your expertise and skills in this cutting-edge field.
Building Multimodal Chat Application on AI PC
This hands-on-lab offers a comprehensive guide on packaging an efficient multimodal chat application for an AI PC using PyTorch. Participants will dive into the intricacies of developing a chat application that leverages text, image, and potentially other modes of interaction, all powered by the local capabilities of Intel® Core™ Ultra Processors. From the initial development stages to the final deployment steps, attendees will gain hands-on experience and valuable insights into selecting the right model and creating an efficient multimodal chat application. This session is ideal for developers looking to expand their expertise with local GenAI and PyTorch on AI PCs. Additionally, this hands-on lab contributes to the requirements for the digital certificate of competency titled ""Essentials for Building AI PC Applications,"" further validating your skills and proficiency in this advanced domain.
Building Safe and Compliant Gen AI Apps with Prediction Guard
This session tackles key challenges in developing multi-modal Large Language Models (LLMs) powered applications, focusing on preventing hallucinations, ensuring data privacy, and maintaining compliance. Participants will learn to build robust Gen AI applications using LLM APIs, incorporating advanced safeguards and governance mechanisms on the Intel® Tiber® Developer Cloud. Through practical exercises and expert guidance, developers will gain insights into creating secure, compliant, and efficient AI applications, ready for deployment in sensitive and regulated environments.
Cognitive AI: Commonsense Reasoning of Multimodal Agents
Discover how human-centric, cognitive AI is the future of machine learning. By 2025, machines are expected to advance in understanding language, integrating commonsense knowledge, reasoning and autonomously adapting to new circumstances. A key tenet of this evolution is multimodal cognition — machines will gain knowledge from a variety of inputs to understand and apply reasoning like humans. Multimodal cognition will bring machines closer to human-level performance in a variety of real-world applications. Learn how Intel® Gaudi® 3 AI accelerators are built to handle demanding training and inference for multimodal AI. Watch a demo through Intel® Tiber™ Developer Cloud on how to save time and power.
Compute Express Link™ (CXL™) in Datacenters Today
Since its inception in 2019, Compute Express Link™ (CXL™) protocol has evolved from v1.1 to v3.1 in a short span of just four years. However, the CPU & device manufactures will lag in terms of CPU/device capabilities compared to the spec version. This panel will examine where we are today & where we will go in near future.
Confidential and Trustworthy AI, Powered by Intel
Gen AI is top of mind for many enterprise developers. But how do we keep these deployments secure? In this session, you'll learn how to use Intel technologies to promote Confidential and Trustworthy AI within an enterprise environment. We will explore scenarios and best practices geared towards securing your data for best Gen AI outcome.
Confidential Computing: Securing Generative AI
As Generative AI (GenAI) gains traction, security concerns about sensitive models and data may hinder business transformation. Confidential computing offers solutions to AI security challenges with isolation and verification while helping address concerns around privacy, provenance, and access control for models, parameters, and data. Experts from Intel and Google Cloud will demonstrate how confidential computing secures high-value GenAI models and data. Learn about algorithmic innovations to enhance trust in GenAI and deploying performant models on Google Cloud Platform using Intel® Trust Domain Extensions (Intel® TDX).
Cost Efficiency for IO Intensive AI
Intel and enterprise customer Pinterest will co-present how a cloud-native IO-aware scheduling and isolation solution enhances the cost efficiency and scalability of AI. The software stack we developed solves enterprise customers' pain point of low resource utilization and fills a functionality gap in Kubernetes. We will talk about how Intel developed software, which is released under an Intel driver-like license, enables resource sharing, isolation, cost efficiency, and scalability in their cloud infrastructure.
Deep Dive: What's Next for AI Kernel Frameworks on Intel GPUs
The open source AI software ecosystem has grown rapidly in the past several years. In particular, programming languages such as OpenAI Triton and Nvidia CUTLASS are becoming increasingly popular for developers who demand performance for custom kernels. Come hear from Intel® engineers who are implementing the next generation of AI programming languages on Intel® GPUs and enabling a simplified developer experience across vendors.
Deliver Quantum safe secure VPN tunnels with PCIe card
Quantum computers when available will break most public-key cryptography in use today. As governments develop standards for post-quantum cryptography (PQC), RFC 8784-compliant solutions can be applied to provide PQC to networks using symmetric key crypto. Arqit delivers an RFC 8784-compliant solution, which can be deployed at the edge using PCIe card based on Intel® NetSec Accelerator Reference Design. This solution hardens and automate symmetric key negotiations to deploy keys in rapid succession to the existing network infrastructure. Learn how Arqit designed its PQC solution and how you can use the Intel NetSec Accelerator Reference Design to implement and scale security applications.
DirectML: A High-Performance Machine Learning Platform for Windows
DirectML is an API that unlocks the potential of machine learning (ML) on any device that supports DirectX 12. It offers a low-level, hardware-agnostic interface for ML inference and training, harnessing the power and efficiency of modern hardware acceleration. In this presentation, we will share the vision, architecture, and discuss current development of DirectML and related technologies from GPU to NPU, ONNX to PyTorch, and native to the web, and demonstrate how it can speed up various ML tasks, such as computer vision, natural language processing, and generative AI.
Distributed Ranges: C++ Library for Multi-GPU Programming
Find out how Intel’s Distributed Ranges productivity library uses standard C++ to automatically parallelize users' code across multiple GPUs and nodes. It offers a collection of data structures, views, and algorithms for building generic abstractions and provides interoperability with MPI, SHMEM, SYCL and OpenMP, and portability on CPUs and GPUs. NUMA-aware allocators and distributed data structures facilitate development of C++ applications on heterogeneous nodes with multiple devices, and achieve excellent performance and parallel scalability by exploiting local compute and data access. Learn about collaboration opportunities and provide feedback on ways to extend the library for developers.
Edge AI - Cloud-Native DevSecOps in the Age of Edge AI
We see no signs of slowing down in the growth of the internet. With the AI and Web 3.0, cloud computing plays a big role in improving the lives of many through on-demand access and processing of data by AI-powered devices such as mobile phones, medical devices, IoTs, and software-driven vehicles. And in order to achieve this unprecedented rate of data being processed on the edge, the need for an integrated, cloud-native DevSecOps is essential for the secure development and deployment of these Edge AI devices. Today, we will explore strategies to securely develop, test, deploy and operate lightweight AI models on edge devices. We will show the value of cloud-native test management systems and how it can build and deploy robust Edge AI devices. Finally, we will discuss how digital feedback loop provides a single pane of glass for operators of these Edge AI devices to make data-driven decisions in the continuous improvement and performance of these devices.
Edge AI for Retail, Powered by Intel Tiber
Enterprise developers who work in retail environments face a daunting task of managing diverse Edge use cases, from inventory management to loss prevention to loyalty programs. Using Intel Tiber technology, we will demonstrate use cases using Computer Vision to execute complex retail operations. While the focus of the session will be on Retail, anyone is welcomed as our use cases can be ported to other industries such as manufacturing, healthcare and beyond.
Edge Native AI on Day 0 and Beyond
The edge is fragmented, dominated by verttical bespoke solutions running on specialized hardware. That’s why Intel has built a new edge-native platform for distributed computing that accelerates AI at the Edge on an open systems basis. In this session, we will take a deep dive into Intel Tiber Edge Platform and how our partners are working with different ITEP elements to deliver business solutions with lower TCO.
EdgeRunner: A Secure, Private, and Local AI Copilot
EdgeRunner offers a revolutionary AI solution bringing an AI copilot experience to any Intel client device. Leveraging the power of Intel’s Lunar Lake NPU, users can run LLMs locally with improved performance. Data stays local giving users enhanced privacy and security with no internet connection required. Users may leverage their personal data on device with a document-agnostic RAG solution, or through LoRAs (Low Rank Adaptors) to tailor the models to their domain. The solution features pair programming capabilities in 16 languages, enabling local code generation and natural language Q&A. The platform provides an on-device OpenAI-compatible API for easy application integration.
Efficient LLM Inference with SqueezeLLM and KVQuant
Quantization is an effective approach for enabling efficient LLM inference; however, it is challenging to quantize LLMs to reduced precision without incurring unacceptable accuracy loss. We present SqueezeLLM, a post-training quantization framework that employs non-uniform quantization to provide a more efficient representation in order to accurately quantize LLM weights, and which isolates outliers in LLM weights and stores them in a compact sparse representation to improve quantization performance. Additionally, we present KVQuant, which accurately quantizes KV cache activations to low precision in order to enable long context length applications.
Empowering AIPCs with Intel's AI Software Stack
Discover how Intel's AIPC software stack is revolutionizing AI development and deployment, offering a suite of tools and frameworks that enable seamless model deployment and an out-of-the-box AI PC experience. Learn the end-to-end model deployment flow, from optimization using quantization and profiling tools to deployment runtimes and frameworks like OpenVINO™, ONNX Runtime, WebNN, and PyTorch. Leverage the AI PC Development Kit for experimentation and rapid prototyping, and utilize the Intel® AI Assistant Builder to create intuitive Generative AI applications. This session empowers you to harness the full potential of AI on PCs, driving innovation for developers, OEMs, ISVs, and OSVs.
Empowering Enterprises with Intel® Gaudi®. Real-world Applications & Success Stories
Join us for a riveting panel discussion where our esteemed customers and partners share their experiences with the Intel® Gaudi® AI Accelerator. This interactive session will provide invaluable insights into the practical applications, challenges, and triumphs encountered while integrating Gaudi into their diverse AI workflows. Hear firsthand how Intel's commitment to open software and scalable solutions has empowered these customers and partners to leverage the full potential of Gen AI. This session will highlight the versatility and efficiency of the Intel® Gaudi® Gaudi 3 AI Accelerator and underscore Intel's dedication to fostering a collaborative and innovative AI ecosystem.
Encrypted Computing: Use Cases, Software & Hardware for FHE
Learn how Encrypted Computing uses full homomorphic encryption (FHE) techniques to enable third-party computation on encrypted datasets without decryption, using only public keys. Intel's strategic roadmap is making FHE technology cost-effective and usable, paving the way for enterprise adoption where data privacy is paramount. Learn more about critical use cases and benchmarks. This talk will introduce a path to an innovative suite of scalable hardware accelerators with a comprehensive software toolkit set. Find out about the early engagement program.
Federated AI: Insights from Early Adopters
This roundtable convenes industry experts from Intel, Siemens, GE, and Swift to discuss trust (and problems with it!) in relation to data and infrastructure. Participants will talk about their very own implementations made possible by Federated AI.
Federated Learning Simplified
The lack of access to private & proprietary data severely restricts AI model accuracy and efficacy. Techniques such as de-anonymization are complex and error prone and further hindered by regulatory restrictions or latency challenges with moving large data volumes. Secure federated AI enables data owners to maintain control over their datasets while collaborating in a federated manner to train and validate models without centralizing sensitive information. This hand-on lab will allow the developer to experience for themselves how they can train or validate a CNN or foundation model securely on a federated architecture.
Fine-tuning and Inference with Intel® Gaudi® AI Accelerators
This hands-on lab offers practical instruction with Intel® Gaudi® AI accelerators, fine-tuning LLMs, and optimizing inference techniques. The session covers hyperparameter selection, model development best practices, and using resources like Hugging Face and PyTorch, serving as an excellent "Gaudi 101" primer for developers new to the Gaudi platform. It also explores techniques for scalable training and low-latency/high-throughput inference, considering key aspects like trade-offs between batch sizes and latency. We will also delve into opportunities for accelerated LLM experimentation across permutations of data and parameters to yield performant models. For developers currently working on GPU architectures, an introduction to GPU migration to Gaudi using the GPU Migration Toolkit will be provided. Additionally, this hands-on lab contributes to the requirements for the digital certificate of competency titled "Essentials for GenAI in the Datacenter," further validating your expertise in this cutting-edge field.
Fireside Chat: Award-Winning Academics on AI and Security
Engage in our fireside chat featuring global academic leaders sharing forward-looking predictions and trends on AI and Security. Discuss significant challenges faced by both private and public sectors and discover cutting-edge research in progress. We are honored to welcome the Intel Hardware Security Academic Award (IHSAA) 2024 luminaries, highlighting impactful research. Join us for an invigorating conversation with IHSAA winners on “AI for Security” and “Security for AI.” This session will explore real-world security challenges and reveal academic advancements that help Intel and our customers fortify their solutions.
GenAI building blocks, blueprints, and enterprise readiness
In 2024, enterprises are expanding AI use cases and transitioning workloads from experimentation to production. Common building blocks, architectures, and testing patterns are being developed and formalized (e.g., by the LF AI & Data Foundation with their Open Platform for Enterprise AI, or OPEA). Leaving this workshop, you will be equipped with processes and knowledge to architect and build enterprise AI apps. Participants will learn about and get hands-on with building blocks for state-of-the-art GenAI systems. They will also implement end-to-end workflows following the most impactful industry blueprints for RAG, summarization, multimodal chat, and coding assistance.
Generating Synthetic Visual Data with Generative AI
Quality data is the foundation of successful machine learning models. AI practitioners often face the challenge of acquiring diverse, representative datasets that can enhance model performance and fairness. This workshop will teach you how to leverage generative AI to create high-quality synthetic visual data. We'll cover models, methods, practical details, and evaluation metrics. We'll begin by motivating the problem, focusing on the limitations of traditional augmentation techniques and simulation engines. Then, we will explore how generative models like diffusion models can help us improve diversity for underrepresented classes. We will delve into advanced techniques for controlled generation. Using tools like ControlNet and latent-space image editing, you'll learn to generate new scenarios, from different weather conditions to various times of day, significantly enriching your training datasets. Finally, we will address outstanding challenges in synthetic data generation. By the end of this workshop, you'll be equipped with powerful tools and techniques to build even more robust and performant predictive models. All models will be run on Intel AI PCs, putting the power of state-of-the-art generative AI at your fingertips.
Generative AI Goes Local on the AI PC
Join a panel with 5 founders of AI startups showing how they built local-first generative AI solutions on the AI PC (including: PreditionGuard, LM Studio, jan.ai, StableEdge, ElectricSQL)
Get Hands on with Intel® Tiber™ Edge Platform
A hands-on workshop for developers to build, manage and deploy an edge-native application with cloud-like ease. The workshop will walk through building 2-3 different edge AI use cases using Intel Tiber Edge Platform components, onboard a local infrastructure into an orchestrator, and deploy applications across with a centrally managed orchestrator.
Implementing AI in vertical industries
Vertical industries are undergoing a digital transformation that challenges them to harness vast amounts of data with AI. Edge-enabled AI allows industries to learn from the data and close the loop to make in-the-moment, informed decisions that allow operations to eventually run autonomously. Intel’s enterprise-ready modular solutions help vertical industries quickly develop and deploy edge AI solutions. These modular SW blueprints pave an easy path to accelerated, highly performant and cost-effective edge AI. In this session learn how Intel provides validated AI solutions across verticals including Manufacturing, Retail, and more.
Improve Device Management in Kubernetes
Dynamic Resource Allocation (DRA) is a new API spearheaded by Intel and community developers for managing devices like accelerators in Kubernetes. This presentation will give an introduction why we drive this Kubernetes framework overhaul and how it will benefit AI/ML workloads to run efficiently on Kubernetes.
Improve Visual Quality of AI in Graphics, Gaming & Streaming
Discover how Intel Labs is working on new ways to identify flaws that can affect the visual quality of AI-generated content in gaming and streaming. Advanced rendering techniques like Intel's neural supersampling (XeSS) and real-time global illumination (RTGI) have greatly improved the look and performance of video games. However, they can also create unwanted effects like ghosting, flicker, blur, and jagged edges. Using spatio-temporal quality metrics, developers can determine whether these issues are noticeable to players and how they affect the overall gaming experience.
Improving Ceph economics with QAT hardware offload
Ceph, the words most popular open source software defined storage system, has offered storage efficiency features such as block device compression, object compression and server-side object encryption for a number of releases. However, enabling these features has always come as a trade-off between the additional performance required (in terms of cores/GHz) vs the raw storage cost, ultimately driving users away from these features. In this talk we will walk through several different scenarios where Intel's QAT offload is used to enable these features without significant overhead to primary processing, and still yields greater performance without causing increased cost per GB.
Innovating with LLMs on Venturely.io: The 'Expedite' Case
This session explores the technical use of LLMs for business model innovation on Venturely.io, exemplified by the European project 'Expedite,' which implements digital twins for energy communities. We will showcase how open-source LLMs, combined with a Retrieval-Augmented Generation (RAG) setup and a business model database, can support the development and rollout of new business models. Using 'Expedite' as a case study, we demonstrate selecting, fine-tuning, and deploying AI models. The presentation also covers infrastructure considerations and participant usage experiences.
Intel® Xeon® 6: Powering Next Wave AI, HPC, Data Services
Delve into the new cutting-edge Intel® Xeon® 6 processor with Performance-cores, a game-changer in AI, HPC, and data services workloads. Uncover distinct features, such as Intel® AMX, MRDIMM and embedded IO accelerators, and how they help drive faster business outcomes, enhanced efficiency, and unparalleled user experiences.

