Agenda

A huge promise of LLM-powered agents is to not only provide simple Q&A search interfaces, but to automate 50%+ of operational work that a knowledge worker typically does. A lot of the day-to-day of a knowledge worker - whether an investment analyst, customer agent, or engineer - is ingesting a large volume of unstructured data and using that to synthesize new insights or take actions. This talk walks through the e2e architecture for building production knowledge agents that can actually do useful work over your data. This requires both a production-quality knowledge management layer as well as agent architectures that can reliably solve certain use cases.

Discover how UiPath's Platform can enhance automation efficiency through agentic orchestration, an approach that seamlessly connects AI agents, robots and humans. This session will explore the integration of UiPath Agent Builder and Agentic Orchestrator to manage complex automation processes across multiple systems. Attendees will gain insights into connecting different automation tasks and improving workflow execution.

Resolve acts as an AI site reliability engineer that helps the lives of on-call engineers. We autonomously troubleshoot and resolve production issues and handle operational tasks, dramatically reducing MTTR and freeing up engineers to focus on building. By combining a deep understanding of production environments with state-of-the-art agentic AI, Resolve aims to accelerate how engineers build, deploy, and maintain real-world software systems.

This workshop will guide participants through designing complex, multi-step workflows using AI agents that leverage Julep AI's persistent memory and task management capabilities. Attendees will learn how to integrate external APIs and services using built-in tools to enhance task automation and efficiency. We will also cover strategies for reliable workflow execution with self-healing features, ensuring robust task completion.

This workshop explores how AI Coding Assistants like Sweep work in the IDE. Learn how to leverage AI through your favorite IDEs like IntelliJ, PyCharm, and Android Studio. Discover best practices for using AI to streamline your day-to-day coding workflows.

MCP redefines how enterprises can operationalize AI by addressing critical gaps in traditional tool integration. In this workshop, we’ll explore how to implement agents that take advantage of MCP’s capabilities. One key benefit is dynamic tool discovery—unlike static frameworks that require manual tool wiring, MCP enables agents to discover and utilize new tools at runtime, allowing organizations to quickly change how the tools are implemented without rearchitecting their systems. Centralized governance further streamlines operations by managing updates, permissions, and compliance policies from a single MCP server, ensuring consistency across all agents and eliminating version-lock issues. Finally, MCP offers elastic scalability, enabling tools to scale independently of the agent architecture so enterprises can grow their systems without disrupting existing infrastructure. Beyond implementation, we’ll also dive into the core concepts behind MCP and discuss why it’s gaining momentum across the AI landscape.

Databricks Assistant is a compound AI agent tailored to Data Scientists, Data Engineers, and SQL Analysts. In this session, I'll explore some of the lessons we learned from building the Assistant along with tips for applying AI to enterprise data. Come learn about UX best practices, Text-to-SQL, fine-tuned code models, and how best to evaluate quality.

Join an AI expert from Meta for an in depth look at the latest advancements in the open generative AI stack. This session will cover from the metal to the agent how large scale AI systems are built, the tools used and how you can build your own with open source AI from Meta including PyTorch, Triton, Llama and more.

Join us to explore how multi-agent orchestration can streamline complex workflows by coordinating AI agents across diverse tasks. Learn to simplify development processes with easy planning, building, and monitoring features, while ensuring enterprise-readiness and scalability. We'll demonstrate universal integration across platforms to optimize AI implementation for real-world applications.

This session explores how to create advanced AI agent workflows using AssemblyAI's Streaming Speech-to-Text technology, enhancing the responsiveness and accuracy of real-time voice applications. Participants will learn to leverage high-accuracy, low-latency capabilities to improve voice-driven system efficiency and user experience. We'll walk through how you can add voice as a medium to any application and use it to perform actions on behalf of the end user.

In this workshop, participants will explore how to create hyper-realistic AI agents that engage users in seamless, real-time conversations. Learn to utilize Tavus.io's API-first design for smooth integration and management of AI video experiences. Discover how empathetic visual perception and multilingual support amplify engagement and accessibility for diverse user bases.

As enterprise adoption of AI accelerates, the need for reliable, auditable, and adaptive systems has never been greater. In this hands-on session, Cyril Gorlla, CEO of CTGT, reveals how leading companies are gaining real-time visibility and control over their models in production — without retraining, redeployment, or going offline​. You’ll learn how to detect hidden bias, censorship, or hallucinations in both open-source and proprietary models, and apply CTGT’s feature-level patches to dynamically restore accuracy and transparency. CTGT’s platform uniquely exposes a model’s internal understanding of concepts and fixes failure modes on the fly, instead of relying on separate “judge” AIs or costly re-training​. Whether you’re building GenAI products or evaluating AI systems for high‑stakes use (finance, healthcare, national security), this workshop will show you how to make AI work reliably, ethically, and safely in any environment.

In this workshop, we'll explore how Sensfix's Multimodal Agentic AI is revolutionizing facilities maintenance by autonomously managing tasks across various modalities—text, images, audio, and video. Attendees will gain insights into integrating AI-driven workflows to enhance operational efficiency, reduce downtime, and proactively address maintenance issues. Through real-world examples, we'll demonstrate the practical applications of Agentic AI in transforming traditional maintenance processes.

This session explores how Armada's Edge Computing and OpsAI solutions revolutionize safety and compliance by providing real-time monitoring and incident management in challenging environments. Discover how AI enhances safety protocols with features like PPE Compliance Monitoring, Zone Monitoring, and Trip, Slip & Fall and Fire & Smoke Detections. Learn how integrating AI with edge computing enables proactive safety measures and efficient response strategies for software engineers and AI experts.

Discover how to achieve GPT-4 quality results at a fraction of the cost with Predibase's approach to fine-tuning small language models. Learn about serverless deployments, the flexibility of using your own VPC, and leveraging open-source technologies such as LoRAX and Ludwig. This session will demonstrate customizable strategies to improve your AI workflows while maintaining control over your data and infrastructure.

This session will explore how Sapien creates an unique understanding of each company through library learning. We bring verticalized intelligence while keeping our underlying code general, allowing us to scale to new companies and industries quickly.

This hands-on workshop will guide you through building a voice-enabled AI support agent that seamlessly connects to your tools and knowledge base. You’ll learn how to integrate real-time voice interactions, automate customer inquiries, and enhance support efficiency using AI-driven workflows. By the end, you’ll have a fully functional voice agent capable of delivering personalized, context-aware responses across multiple channels.

Have you built a text or image search demo with a vector database, but still feel like you have a long way to go to enable a core user flow like natural language search, real-time personalization or retrieval for your shopping assistant? Are you dealing with complex data like job listings, travel destinations or second hand fashion items with 10s of metadata properties? In this workshop we will build a vector search-powered system that can deal with complex data and power highly relevant & well converting natural language search with queries like "vintage sweaters under 200" or "cozy highly rated hotels near manhattan midtown with good wifi". We will personalize these results and build product discovery with a real-time feedback loop using click data. We will learn how a fashion retailer added $15M of incremental revenue using Superlinked's ability to create custom embeddings that fit *all* their 1P data signals.

We've all been there—that moment when a brilliant idea strikes, but the path from concept to actual product seems overwhelming. Whether you're a startup founder, a product manager, or a developer with a vision, turning an idea into reality can feel like climbing a mountain. That's where Trae comes in—your new secret weapon for bringing ideas to life. This workshop will show you how an AI-native IDE can break down the barriers between your innovative concept and a real working prototype. We'll demonstrate how cutting-edge AI coding tools can help you sketch out your idea, quickly build core features, and transform your vision from a napkin sketch to a functional product

Discover how agentic workflows can revolutionize software development with Qodo AI-powered tools. Learn to utilize context-awareness and automated review workflows to ensure impeccable code integrity. This session will explore real-world applications, focusing on best practice integration and AI-driven insights to optimize your development process.

This session will cover Sama's comprehensive Model Evaluation services, emphasizing the importance of aligning AI systems with ethical guidelines and improving their performance. Attendees will learn how to systematically assess AI outputs, identify inaccuracies and vulnerabilities, and employ strategies for accelerated time-to-market with robust model validation. Leveraging Sama's expert-driven platform, participants will discover how to build reliable, high-performing generative AI models while adhering to industry standards.

In this workshop, we’ll first explore the common challenges developers face when working with large codebases, especially when traditional code assistants struggle to provide context for complex projects. Understanding the relationships between files, functions, and dependencies is crucial, but most tools fail to capture this structure, leaving developers with limited insights. Next, we’ll dive into solving this problem by using CodeGPT to load your own code repository. We’ll visualize how the platform builds a Knowledge Graph from your code, transforming it into a structured base of knowledge. This graph will then be used throughout the software development process, from generating code to verifying tasks, providing smarter and more accurate insights to enhance your workflow.

In this interactive session, we'll go through real-world strategies for integrating AI tools into a realistic Product Development Life Cycle (PDLC)—from idea validation through product iteration. Learn how to quickly evaluate and prototype concepts with V0 and Replit, creating shared assets both developers and product can iterate upon, integrate analytics to rapidly capture user insights. Explore practical planning with Zentrik in realistic team scenarios, and empower your product and design teams by effectively deploying Cursor with product-specific data for accelerated iteration.

In this session I will give a walkthrough about current status of APIs and how developers use them, the challenges,trends and how Swytchcode solves the problem. I will keep this session interactive with attendee participation.

Building a prototype AI application is easy, but getting it to work reliably in production is not. Just ask the developers of New York’s MyCity, whose AI got caught telling businesses to break the law. (And there are many other stories like this!) We envision a future where AI applications avoid such issues by design, using automated detection and remediation of bad responses. In this hands-on workshop, we’ll walk you through a case study of building a reliable retrieval-augmented generation (RAG) application. We’ll start by implementing a baseline application, then integrate Cleanlab’s automated real-time evaluations to detect general issues such as knowledge gaps and hallucinations as well as custom application-specific evaluation criteria, and finally add real-time remediation of these issues.

Hybrid search, filtering, re-ranking and text2sql offer a limited way to integrate metadata into your vector search system. Wouldn't it be better if your embedding model understood both your unstructured data AND the associated structured metadata? Learn how a fashion e-com company solved this problem across search, recommendations and agentic retrieval use-cases with Superlinked - an open source encoder stacking framework.

As the AI industry shifts from centralized cloud models to edge-native intelligence, a new era is emerging—one where powerful AI models can be trained, fine-tuned, and deployed directly on personal devices. This panel, featuring industry pioneers Hina Dixit (CEO, Decompute), David Kanter (Founder, MLCommons), and Aishwarya Srinivasan (AI Leader, Fireworks AI), explores the technological breakthroughs, infrastructure evolution, and ethical imperatives of decentralizing AI. Together, they’ll discuss how on-device AI is unlocking new frontiers for privacy, autonomy, and performance. This session will uncover how decentralized AI empowers individuals and developers alike, how open standards and benchmarks are shaping performance at the edge, and why the next generation of AI will be personalized, private, and truly yours.