Toronto Data Engineering Meetup

An AI agent can name the right root cause for the wrong reason. For the Toronto Data Engineering Meetup, Rootly and ClickHouse brought the city’s data community to our HQ to look underneath an incident investigation—the data agents query, the traces that show how they reached an answer, and the signals that turn symptoms into a diagnosis—with talks from:
- Nischay Dhiman — Senior AI Engineer at Rootly
- Vishal Jethnani — Senior Director of SRE at Loblaw
- Jovin Manjaly — AI Product Manager at Caseware
Nischay opened with a phone buzzing at 3:00 a.m. and a replay system for testing incident agents: once a cause is confirmed, logs, spans, metrics, and events are copied from Datadog into ClickHouse, sorted so each replay reads one contiguous incident slice. The agent still calls a Datadog-style search while ClickHouse answers underneath. He also covered the agent that had the answer key: right answer, wrong reason.
Vishal described enterprise observability as a nervous system, in which applications and infrastructure generate vital signs, collectors move them across the business, and correlated evidence turns symptoms into a diagnosis.
Jovin then showed what trace spans reveal when an agent’s output looks fine: soft failures, config drift, and slow or serial turns. Audience questions followed the talks.















