We are seeking an engineering-focused

Data Scientist

to build, operationalize, and maintain production-grade retail forecasting and optimization models. In this role, model development goes hand-in-hand with operational reliability: success is measured by accurate forecasts and stable, low-latency, deterministic pipelines running natively on the Databricks Lakehouse. You will bridge the gap between applied data science and machine learning engineering. Working in close partnership with

Product

and

Data Engineering , you will own the operational lifecycle of item-store level demand forecasts and downstream replenishment/optimization engines—from distributed feature pipelines to automated Databricks workflows, model tracking, and runtime monitoring. Key Responsibilities

Production Forecasting & Optimization:

Develop, calibrate, and tune time-series forecasting engines and downstream supply chain/inventory optimization logic using tree-based ensembles ( LightGBM ,

CatBoost ) and distributed Python/PySpark. Databricks-Native MLOps:

Build, schedule, and maintain automated training and batch inference pipelines using

Databricks Workflows

and

Jobs . Leverage

MLflow

for robust experiment tracking, artifact logging, and model registry management. Cross-Functional Partnership: With Product:

Translate business requirements into technical specs, define operational SLAs, and provide technical feasibility assessments for new forecasting features. With Data Engineering:

Establish strict data contracts, define schema validations, and optimize data ingestion/consumption patterns from upstream Delta tables. Pipeline Quality & Stability:

Treat ML pipelines as critical production software. Implement pre-inference data validation gates (e.g., schema checks, missingness thresholds, null checks)



and automated alerting to prevent corrupted data from reaching scoring jobs. Model & Pipeline Observability:

Track pipeline health, monitor runtime performance, and detect feature drift, target drift, and forecast degradation across high-cardinality retail catalogs. Software Excellence:

Write modular, maintainable, and testable code. Champion version control best practices, unit/integration testing with pytest, and automated CI/CD checks within Git. Spec-Driven Execution:

Embrace a Spec-Driven Development (SDD) mindset, leveraging modern agentic AI development workflows (e.g., Cursor, Claude Code) to move rapidly from research to reliable production code. Required Technical Skills

Tree-Based Ensembles:

Hands-on experience developing, tuning, and deploying gradient boosted decision trees—specifically

LightGBM

and

CatBoost —on high-cardinality, tabular, and time-series datasets. Time-Series Retail Forecasting:

Deep practical understanding of demand forecasting challenges: trend, seasonality, calendar events, promotional uplifts, stockouts, and intermittent/sparse demand patterns. Databricks Platform:

Proven experience building within the Databricks ecosystem, specifically authoring and managing multi-task

Databricks Workflows/Jobs , navigating Delta Lake, and using

MLflow

across the model lifecycle. Data Manipulation & PySpark:

Solid proficiency in Python and PySpark for distributed data processing,



feature engineering, and memory-conscious transformations across massive retail datasets. Software Engineering Fundamentals:

Solid understanding of clean code principles, modular package design, virtual environments, automated testing (pytest), and standard Git workflows (pull requests, branching, code reviews). Preferred Qualifications

Optimization & Supply Chain:

Familiarity with inventory optimization mechanics (safety stock calculation, reorder point modeling, lead time variability, allocation constraints). Databricks Advanced Features:

Experience leveraging Delta Live Tables (DLT), Unity Catalog for data and model governance, or Photon compute engine. Explainable AI ( XAI ):

Experience implementing TreeSHAP or similar interpretability methods within production batch scoring jobs. Continuous Integration:

Experience setting up or integrating with CI/CD pipelines (e.g., GitHub Actions) to automate testing and deployment into Databricks workspaces. Cultural Alignment & Values

Production Mindset:

You believe a model is only finished when it is tested, automated, monitored, and running reliably in production. Ownership & Root-Cause Thinking:

When a pipeline fails or a metric degrades, you dig into the logs, identify the root cause, write a regression test, and implement a durable fix. Collaborative Communicator:

You easily speak the language of business trade-offs with Product managers and system architecture with Data Engineers. The estimated pay ranges for this role are as follows: $120,000 - 140,000 CAD The successful candidate’s starting salary will be determined based on permissible, non-discriminatory factors such as skills, experience, and geographic location.

#J-18808-Ljbffr

📌 Data Scientist (Winnipeg)
🏢 Upshop
📍 Winnipeg