Transport · Montréal
Canadian operator: 12 data silos → lakehouse, real-time KPIs.
Inconsistent KPIs, dashboards 48h late. Databricks lakehouse, medallion, dbt, self-service BI.
KPI
60%
autonomous analysts
Duration
9 mois
Team
6 engineers
Hub(s)
Montréal
When every department gives you a different number for the same KPI, you're not making decisions: you're arbitrating between opinions.
The context
Canadian transport operator, 11,000 employees, Montreal hub. 12 historical data silos (operations, HR, finance, ticketing, maintenance, etc.), aging Oracle data warehouse, Excel dashboards sent by email.
The problem
- 12 data silos with no shared governance
- Inconsistent KPIs across departments (up to 18% gap on the same indicator)
- 48h late dashboards, manual updates
- No data catalog, duplicates and ambiguous definitions
- Analysts stuck on SQL extracts, low business autonomy
The approach
Databricks data lakehouse with medallion architecture (bronze/silver/gold), Unity Catalog governance, versioned dbt transformations, self-service Tableau BI with semantic layer.
The pillars
- Real-time ingestion via Auto Loader (Kafka + files)
- dbt dimensional modeling, mandatory quality tests
- Unity Catalog for governance, lineage, RBAC
- Semantic layer exposed to Tableau (centralized business definitions)
- Enablement program: 60% of analysts trained in 4 months
The stack
- Databricks Lakehouse Platform on Azure
- dbt Cloud for versioned transformations
- Unity Catalog for governance and lineage
- Tableau Cloud with semantic layer
- Apache Airflow for ingestion orchestration
The results
- Single source of truth: 100% of operational KPIs reconciled
- Dashboard latency: 48h to real-time (sub-minute on 80% of KPIs)
- Autonomous analysts: 60% in 4 months (vs 12 targeted)
- Data costs: -22% despite 3x on processed volumes
- Data quality issues: -76% in 9 months
« For the first time in 15 years, my operations and finance teams argue about action levers, no longer about numbers. That's the ROI of a data platform. »
What we learned
Unity Catalog is Databricks' real differentiator, not the Spark engine. dbt scales very well up to 800 models, beyond that you need to invest in modularization. Mistake: we shipped the gold layer before consolidating silver, expensive rollbacks. To redo: never open analyst access before 90% of dbt tests are green. Otherwise, you lose trust and don't get it back.
// Read next
Luxury jewellery & watchmaking · Genève + Paris + Tokyo
Cartier: from audit to in-house private LLM.
Compass (front + back architecture audits), Mapper (watchmaking + jewellery product generator), competitive data ETL on BigQuery, and now a private LLM fine-tuned on Cartier's own infra. A long-term tech partnership on the data and AI stack of a luxury house.
LLM privé
fine-tuned on Cartier infra
Tier-1 bank · Paris
BNP Paribas: Reference Book PO, from React/Redux to product AI agents.
Three Abbeal engineers at the core of the PO Marketplace. React/Redux/Node platform initially, now augmented with a product RAG, Claude agents for PM assistance, and an event-driven Kafka layer to scale.
RAG
PO product catalog
Digital banking / FinTech · Tokyo (Tamachi)
Money Forward: data backbone of a brand-new digital bank in Tokyo.
Money Forward, a Japanese FinTech leader listed in Tokyo, partnered with a top-tier Japanese banking group to launch a brand-new digital bank built from scratch. Abbeal partners on the Data Engineering side: designing and operating the Data Hub (Databricks + Delta Lake + dbt + AWS Tokyo) serving JFSA reporting, AML, risk management.
Data Hub
from-scratch digital bank Tokyo