From prompt to production: Spec-driven AI development with AWS Kiro

Machine learning models are only as valuable as the software systems that deploy them. In this session, we explore AWS Kira – an agentic IDE that transforms natural-language intent into structured specifications, technical designs, and sequenced implementation tasks before writing a single line of code. Attendees will learn how Kiro’s spec-driven development workflow generates requirements documents, architecture designs, and property-based tests using agents powered by Amazon Bedrock. We will demonstrate how ML practitioners can go from a model prototype to a production-ready application with engineering rigor – including automated edge-case detection, dependency tracking, and parallel agent execution – all within a single IDE. Whether you’re building inference pipelines, model-serving APls, or ML-powered features, this talk shows how agentic engineering replaces ad-hoc coding with maintainable, verifiable software.

Speaker’s profile

Vishesh serves as a Principal Architect within the AWS Professional Services team in Singapore, specialising in Application Development & Architecture, Infrastructure, and Security. He collaborates closely with senior decision-makers and development teams from ASEAN’s largest enterprises, driving cloud adoption strategies, solution optimisation, and robust security implementations. His expertise spans across designing scalable, secure, and high-performance cloud environments, ensuring businesses maximise the value of AWS technologies. Also, Vishesh holds all active AWS certifications, reflecting his deep technical proficiency and commitment to learning to build industry-leading AWS solutions.