The ML Engineer builds the classical retail-ML cores that power the highest-stakes agents on an AI-native retail decisioning platform — demand forecasting that must beat a legacy system, replenishment and allocation models, causal-insight models for executive narratives, and pricing / promotion / markdown / assortment models. The role consumes the enterprise MLOps platform (model registry, drift detection, feature store, library wrappers) and contributes use-case-specific implementations.
Remote candidates outside of Thailand are welcome to apply.
Key Responsibilities:
- Build, train, evaluate, and deploy classical retail ML models — forecasting, replenishment, allocation, causal inference (DoWhy / EconML), pricing elasticity, promotion lift, markdown optimisation, assortment.
- Use company-curated classical ML wrappers (Prophet, statsmodels, DoWhy / EconML, LightFM, scikit-learn, XGBoost, LightGBM) — do not rebuild open-source libraries from scratch.
- Author per-model evaluation methodology appropriate to each model class (forecast MAPE, classification accuracy / precision / recall, causal precision).
- Register every model in the enterprise Model Registry with model cards; configure drift-detection thresholds; use the enterprise Feature Store for shared features.
- Beat a legacy forecasting system by a measurable margin (MAPE improvement) and document evidence for trust-gate progression alongside the legacy run.
- Build causal models for executive-insight agents using DoWhy or EconML; document causal assumptions; ensure mandatory citations for narrative outputs.
- Partner with AI Engineers on ML model ↔ agent integration (invocation contracts, latency budgets, fallback behaviour); co-design HITL gate criteria for ML-heavy agents.
- Partner with Suite Product Owners on BU adoption, gate criteria, success metrics; document per-model business value (forecast accuracy → inventory savings, replen accuracy → stock-out reduction).
Requirements
- Bachelor's or Master's degree in Computer Science, Statistics, Applied Mathematics, or a related discipline.
- 5+ years building production ML systems with retail or commercial decisioning models (forecasting, replenishment, pricing, recommendation, or comparable).
- Strong Python and Spark / PySpark; SQL fluency.
- MLOps consumer experience — has registered models, configured drift, used a feature store.
- Cloud + Databricks (or equivalent lakehouse) production experience; Azure preferred.
- Causal inference exposure (DoWhy / EconML).
- Eval discipline — knows how to design appropriate evals per model class.
- Retail / commerce domain fluency or rapid acquisition.
Preferred Qualifications
- Retail forecasting at multi-store / multi-SKU scale; promotional lift / markdown optimisation in production.
- Causal inference in commercial decisioning; replenishment / allocation algorithms.
- Online learning / near-real-time inference.
- Vendor certifications such as Databricks Machine Learning Professional or Azure AI Engineer Associate
