HALL 3 - Tech Talks

As enterprises rapidly deploy large language models into real-world applications, achieving high throughput and low latency has become a critical requirement for modern AI infrastructure. This session explores what it takes to run LLMs at scale—from optimizing model serving pipelines and distributed compute to efficient workload scheduling and inference acceleration. We’ll take a closer look at how advanced GPU architectures and AI platforms from NVIDIA enable faster processing, reduced response times, and consistent performance even under heavy demand. Attendees will gain insights into practical strategies for designing scalable AI systems, balancing cost and performance, and building infrastructure that can support next-generation generative AI workloads in production environments

As organizations move toward building and operating their own AI infrastructure, managing large-scale GPU environments becomes both an opportunity and a challenge. This session shares real-world lessons from running a production AI stack powered by a 13-GPU fleet—covering how teams design, deploy, and optimize infrastructure to support demanding AI workloads and large language models. From workload orchestration and performance tuning to cost management and system reliability, the discussion will highlight practical insights gained while operating GPU clusters in a live environment. Attendees will learn what it takes to maintain stability, scale efficiently, and ensure consistent performance when running AI systems in production

As Generative AI moves into HR, the bar for employee experience has shifted considerably. Employees expect accurate, context-aware answers and organizations need systems they can actually audit. This workshop presents an end-to-end approach to designing and implementing an HR Copilot using a Hybrid Retrieval-Augmented Generation (RAG) framework with a multi-agent architecture.

Modern AI systems are rapidly evolving from simple prompt-response models to autonomous agents capable of reasoning, using tools, retrieving knowledge, and executing complex workflows. But what actually makes an AI agent work?
In this session, The Anatomy of an AI Agent, we will break down the core building blocks behind modern agentic systems. We will explore how large language models, memory, retrieval, planning, and tool execution come together to create intelligent, reliable, and production-ready AI agents.
The session will focus on practical architecture patterns used in real-world systems, including how agents reason over data, interact with external tools, maintain context, and handle multi-step tasks. Attendees will gain a clear mental model of how AI agents are designed, the trade-offs involved, and what it takes to move from demos to scalable, real-world deployments.
This talk is intended for engineers, architects, and AI practitioners who want to understand how modern AI agents are actually built under the hood.

This talk will explore how Scapia is scaling its CX Bot in production by building robust evaluation frameworks that continuously measure response quality, accuracy, and reliability in real time. The session will also dive into model resiliency, highlighting how Scapia is developing an internal platform that enables employees to easily switch between different models, experiment rapidly, adopt best practices, and share learnings across teams. In addition, it will cover Scapia’s approach to hosting models within its own data center to maintain stronger control over data governance, security, and policy compliance while operating AI systems at scale.

Physical intelligence, the ability of machines to perceive, reason, and act in the physical world remains one of the key frontiers in artificial intelligence. While modern AI systems have made significant progress in language, vision, and pattern recognition, they still struggle to fully understand and interact with the physical laws that govern real-world environments. This session explores the technology gaps in physical intelligence, focusing on the disconnect between current intelligence models and the principles of physics that shape real-world interactions. It will examine limitations such as weak physical reasoning, challenges in predicting object dynamics, and the difficulty of learning from limited real-world data. The talk will also highlight the “physics gaps” in today’s AI models, where purely data-driven approaches fall short in capturing causal and dynamic properties of the physical world and discuss emerging opportunities in areas such as embodied AI, simulation-based learning, robotics, and hybrid physics-AI systems that aim to bridge these gaps.

As large language models move from prototypes into enterprise workflows, teams across the industry increasingly face a practical operational alignment problem: how to steer model behavior toward business outcomes in a reliable and measurable way. At the same time for many practitioners a critical decision remains unclear: whether a use case is best addressed through prompt engineering, supervised fine-tuning, or reinforcement learning. This talk introduces a structured framework for choosing the right approach and a practical method for translating business objectives into reward signals that can be optimized, evaluated, and audited.
Through hands on experiments in incremental preference learning using reward modeling and policy optimization, the session demonstrates how meaningful behavioral shifts can be achieved even with small, carefully curated datasets. It also examines the stability trade-offs between staged training and online RL updates. Additionally, the session distills practical guidance on designing verifiable reward signals from ambiguous business objectives. This segment identifies when exploration and delayed rewards make reinforcement learning necessary and how to avoid common failure modes, such as reward hacking and instability. Rather than focusing on scale alone, the session emphasizes disciplined reward design and systematic experimentation as the foundation for deploying reinforcement learning effectively in enterprise LLM systems.

LLMs today promise endless text generation, but creating high-fidelity synthetic text data that actually reflects complex business logic remains an engineering challenge. This talk moves beyond basic “prompt and pray” techniques to address the various nuances of creating instruction datasets useful for knowledge distillation, domain adaptation, and reinforcement learning (RL) workflows.
We will examine why direct generation often causes datasets to regress to the mean, producing repetitive, safe content that lacks the messy edge cases required for robust training. To solve this, we suggest a systematic, algorithmic approach that treats data generation as an engineering problem. We will discuss how to decompose pipelines into iterative batches to programmatically inject real-world variations. We also conclude with a strategic checklist to evaluate if synthetic data is truly well-suited to your enterprise problem.