Market Data Calendar

Clients consuming TMX real-time feeds run software that reads fields, tags, message types and network channels. When an exchange adds a field, splits a feed or moves a port, their systems may need to change before the effective date.

Those changes arrive as technical notices, trading notices, specifications and announcements, published in different places.

Through client and internal discovery, I mapped how a client discovers, verifies and acts on a change before it reaches their systems.

The failure went beyond clients missing the dates of changes. It also meant missed policy changes, repeated research and different readings of the same notice across teams.

A notice is unstructured text. An event is structured data: the market and feed affected, what changed, the published and effective dates, whether action is required, and a link back to the authoritative source.

That model is what makes the calendar filterable, searchable and personal, which a document library can’t be.

At the core, the product is an AI diff pipeline on AWS Bedrock that compares new and previous source documents and finds what changed. I defined what counts as a material change, the confidence thresholds and the escalation path.

Ambiguous or low-confidence detections go to a person. A wrong event costs trust, and a missed one costs a client’s readiness. I worked with Verity’s engineering team on the document comparison behind it: scanning previous and new agreements and notices for changes that are hard to catch by hand.