About
Over a decade of experience in data, ML and AI working in banking, fintech and info sec. Love engineering challenges and with a transformation technology like AI we are seeing a lot of paradigms changing. Others seem to be pretty robust despite their often prematurely claimed death. Currently working on making agent more reliable by monitoring them in production and providing incident management for agent builders.
Sessions
Agent Fails: A walkthrough of potential agentic failures and how they are different from existing software bugs.
What you will learn:
•Most software fails in predictable ways, e.g. 4XX or 5XXX •Agents fail in implicit ways which are hard to capture using deterministic logic •Understanding expected behaviour and agent logic and finding outliers from that base behaviour is the key to making agents more reliable.