About the engagement
Our client is a large, data-rich manufacturing and distribution business whose data science team built a custom, in-house AI demand forecasting platform. We're conducting an independent assessment of the client's data science capability, and you'll lead the technical evaluation of that forecasting platform. This is an assessment and advisory role, not a build.
Fixed, phased engagement (~160 hours, late October to mid-December 2026) with one concentrated on-site week in mid-November. Industry: large U.S. beverage manufacturing & distribution.
What you'll do
- Perform a hands-on code and statistical review of the client's custom AI forecasting model: codebase, modeling frameworks, feature engineering, pipelines, and deployment architecture.
- Evaluate model performance: forecast error and accuracy, bias, stability, and performance across SKU, pack, customer, and location levels.
- Assess maintenance burden, technical debt, and MLOps maturity (monitoring, retraining, versioning, lineage, feature management).
- Benchmark the current approach against alternative forecasting methods — including promotions, seasonality, events, and hierarchical reconciliation.
- Compare the custom solution against commercial planning platforms and recommend whether to retain, enhance, or replace it.
- Identify true predictive and prescriptive use cases versus BI and standard reporting (inventory and safety-stock optimization, production planning, route optimization, predictive maintenance, pricing, churn, commodity forecasting).
- Provide input on ML workload placement across Snowflake, Databricks, and Microsoft Fabric.
- Work directly with the client's internal data scientist and business stakeholders, translating technical findings into clear, executive-ready recommendations.
- Contribute to the final roadmap, build-vs-buy-vs-partner analysis, and executive readout.
What we're looking for
- 8+ years in data science or applied ML, with a strong focus on time-series and demand forecasting.
- Deep hands-on experience with ARIMA/SARIMA and derivatives, gradient boosting (XGBoost, LightGBM), and deep-learning forecasters (LSTM, Temporal Fusion Transformers, DeepAR).
- Strong Python (pandas, scikit-learn, statsmodels, PyTorch or TensorFlow).
- Proven ability to audit and evaluate an existing ML codebase and production pipeline — not just build new models.
- Solid understanding of MLOps: pipelines, model monitoring, retraining, and experiment tracking (e.g., MLflow).
- Experience in supply chain, CPG, manufacturing, or distribution forecasting.
- Excellent communication skills; comfortable presenting to senior business and technical leaders and working on-site with client stakeholders.
Preferred
- Experience with Integrated Business Planning (IBP) or S&OP: demand, inventory, production, and logistics planning.
- Exposure to commercial planning tools (e.g., Blue Yonder, SAP IBP, Kinaxis).
- Working knowledge of Databricks and/or Snowflake for ML workloads.
- Background in inventory optimization, multi-echelon planning, or route optimization.
- Advanced degree (MS or PhD) in Statistics, Operations Research, Computer Science, or a related field.
Out of scope
LLMs, NLP, and agent development.