The State of GenAI & Machine Learning in 2025

Are your AI developer tools creating a hidden productivity bottleneck? This talk explores the downstream costs of GenAI and the new infrastructure required for building agentic systems.

#1about 5 minutes

The long history and rapid market growth of AI

AI is not a new field, but its market size and user base are growing exponentially, creating significant business opportunities.

#2about 4 minutes

Analyzing the developer productivity funnel for GenAI tools

Most GenAI developer tools focus on the top of the funnel (writing code), creating bottlenecks in testing, operations, and monitoring.

#3about 2 minutes

Overcoming the key challenges of building with GenAI

Adopting GenAI introduces challenges like vendor complexity, moving beyond simple chatbots, and a lack of established best practices for AI systems.

#4about 5 minutes

Deconstructing the modern agentic systems stack

Building robust GenAI requires moving from a model-centric to a system-centric view, encompassing orchestration, data, guardrails, and security.

#5about 3 minutes

Agentic infrastructure and the critical role of data

Effective agentic systems rely on complex infrastructure for non-deterministic data flows and specialized hardware scheduling, underscoring the "garbage in, garbage out" principle.

#6about 2 minutes

New security vulnerabilities and monitoring for AI systems

AI systems introduce unique security risks like data poisoning and require specialized monitoring for performance, explainability, and model drift.

#7about 1 minute

Implementing responsible AI principles by design

To address challenges like algorithmic bias, responsible AI principles must be embedded directly into the design of platforms and infrastructure.

#8about 3 minutes

AI maturity, new roles, and appropriate use cases

AI maturity is a non-linear journey focused on time-to-value, creating new roles like the AI engineer and requiring careful consideration of when not to use GenAI.