AI Developer Day: From Prototype to Production

Chairpersons welcome and opening remarks.

Everyone has an AI proof of concept. Far fewer have AI systems serving millions of requests reliably in production. Engineering leaders discuss what happens after the demo, from LLMOps and infrastructure to observability, developer experience and operational resilience, and the architectural decisions that separate successful AI products from abandoned experiments.

Getting an LLM to answer a question is easy. Building an AI application that remains reliable, observable and maintainable at scale is not. This discussion explores orchestration, evaluation, context engineering, memory and platform design, and why production AI is becoming an engineering discipline in its own right.

Moving generative AI from experimentation to production requires more than powerful models. This session explores practical techniques for optimising performance, managing costs and latency, and building safety and reliability into production-ready GenAI systems. Discover approaches to evaluation, monitoring, guardrails and deployment that help organisations scale generative AI with confidence.

Traditional software testing wasn’t designed for probabilistic systems. Learn how engineering teams are building evaluation pipelines, automated benchmarks and observability frameworks that detect regressions before users experience them to ensure quality and business performance.

As foundation models become increasingly capable, competitive advantage is shifting away from model selection and towards how applications manage context. This session explores the engineering patterns behind reliable AI systems, including context windows, retrieval strategies, memory, tool use and prompt orchestration. Learn how engineering teams are designing context pipelines that improve accuracy, reduce hallucinations and create more predictable AI behaviour in production.

AI agents are moving rapidly from experimentation into enterprise applications, but reliability remains the biggest engineering challenge. Engineering leaders discuss planning, orchestration, tool use, memory and failure recovery to build dependable AI agents

As AI adoption accelerates, the real competitive advantage lies in platform thinking. In this session, Kapil Poreddy discusses how modern AI platforms empower developers to build, deploy, and iterate faster while maintaining governance and reliability at scale. Drawing from experience at Walmart Global Tech, this presentation covers reusable AI infrastructure, internal developer platforms, LLM integration strategies, observability for AI workloads, and balancing innovation with enterprise controls. Attendees will gain actionable insights on building AI ecosystems that enable teams – not just models – to scale.

Successful AI programmes depend on more than powerful models – they require internal platforms that enable developers to experiment, deploy and operate AI safely and efficiently. This session explores the principles behind effective AI platform engineering, covering APIs, reusable components, governance, developer experience and self-service infrastructure that accelerate AI delivery across engineering teams.

Successful AI programmes depend on more than powerful models – they require internal platforms that enable developers to experiment, deploy and operate AI safely and efficiently. This session explores the principles behind effective AI platform engineering, covering APIs, reusable components, governance, developer experience and self-service infrastructure that accelerate AI delivery across engineering teams.