I build a manufacturer's data backwards from the hardest question a customer, an inspector or a director will ask about a single part, and I tell them plainly what the record cannot yet answer.
I am chief data officer of Tessandier Aerospace, a listed component manufacturer in Longueuil, Quebec, with six plants in Quebec, Ontario and Mexico, 4,300 colleagues and C$1.18 billion of revenue in 2025. We machine, heat-treat and inspect engine, airframe, hydraulic and gearbox parts for engine makers, airframers and defence primes. My office of 72 people is responsible for product traceability records, plant and operational data, analytics and AI, and data governance. I report to our chief executive, Benoît Lavallière, and I present to the board's audit and risk committee twice a year.
I came to data through quality. My first job, in 1998, was as a quality engineer at a forge in Trois-Rivières, where a furnace chart that ran out of ink cost a customer four weeks and taught me that a part is only as good as its record. I then ran metrology and quality data at a Longueuil machine shop, built failure records and field-data models for a mining-equipment maker in Sudbury, and joined Tessandier in 2014 to fix the traceability problems an engine customer's audit had found. I became our first chief data officer in January 2022.
The questions I work on are particular to a supplier whose parts fly. How quickly can we name every part made from a suspect lot of titanium? Which inspection and maintenance decisions may a model inform, and which must stay with a certified person? Who owns a plant's data when head office wants comparisons? My answers have cost money, one vendor discount and an argument with our chief operating officer that I would have again.
A recall question answered in hours, not weeks
In 2022 it took us twenty-three days and thirty-one people to tell an engine customer which parts came from fourteen suspect titanium heat lots. In a timed drill in March 2026 we answered a 3,800-part question in three hours and ten minutes, with no corrections afterwards.
A smaller predictive maintenance programme that delivered
I advised against a fleet-wide spindle-prediction rollout whose case claimed C$6.1 million a year. Better failure records and condition monitoring on 64 machining centres cut unplanned spindle downtime by 27% and saved about C$2.2 million a year.
Plant figures that reconcile to the ledger
Six plants counted scrap six ways. A data agreement giving head office 46 shared definitions, and leaving the data with the plants, brought reported scrap within 0.6% of the general ledger, from a gap of 11%.
An inspection model that never accepts a part
Our penetrant-inspection assistant ranks parts and highlights areas for certified inspectors but cannot pass anything. Throughput rose 16%, and customer-detected escapes from that line fell from three in 2023 to none in 2025.
Current focus
I am extending the serial genealogy record to parts made between 2010 and 2015 that remain on active engine programmes, which means reading handwritten amendments on about 260,000 paper traveller pages. I am also rewriting our AI register so that every use touching product acceptance, export-controlled drawings or employee information has a named owner, a stated limit and a review date the audit and risk committee can see.
Conversations I welcome
I am not seeking another executive role. I would consider one non-executive seat on the board of a Canadian manufacturer or regulated-product company outside the aerospace supply chain, with our chief executive's consent and a conflict review. I accept a small number of advisory conversations and speaking invitations on traceability, plant data and AI governance each year.
