Data Scientist/ML Engineer : Demand and Time-Series Forecasting

Experience : 4 - 6 years | Function : Data Science / Machine Learning Engineering

About the Role :

We are looking for a hands-on Data Scientist/ML Engineer with deep expertise in demand and time-series forecasting to own the design, development, and productionization of forecasting systems that directly drive planning, inventory, and business decisions. You will work end to end, from framing the business problem and building models to deploying, monitoring, and continuously improving forecasts at scale.

Key Responsibilities :

- Design, build, and deploy demand forecasting models across multiple horizons (short-term operational to long-term strategic) and granularities (SKU, store/location, region, category).

- Develop and benchmark models spanning classical statistical methods, ML-based approaches, and deep learning architectures; select the right tool for the business problem, not the trendiest one.

- Engineer features from demand drivers such as seasonality, promotions, pricing, holidays and events, weather, cannibalization, and macroeconomic signals.

- Handle real-world forecasting challenges: intermittent or sparse demand, new product (cold-start) forecasting, hierarchical reconciliation, and demand sensing.

- Build robust backtesting and evaluation frameworks with business-aligned metrics (WMAPE, MASE, bias, forecast value added) and rigorous, leakage-free time-based validation.

- Productionize models: automated retraining pipelines, model versioning, drift detection, and forecast monitoring and alerting in collaboration with data engineering and MLOps teams.

- Quantify forecast uncertainty (prediction intervals, probabilistic forecasts) and translate it into actionable inputs for inventory, supply, and capacity planning.

- Partner with business stakeholders (supply chain, planning, finance, commercial teams) to define success metrics, explain model behavior, and drive adoption of forecast outputs.

- Mentor junior data scientists and contribute to team best practices, code standards, and experimentation culture.

Core Skills : Must Have :

Time-Series and Forecasting Expertise :

- Strong grounding in classical methods: ARIMA/SARIMAX, exponential smoothing (ETS/Holt-Winters), and state-space models.

- ML for forecasting: gradient-boosted trees (XGBoost, LightGBM, CatBoost) with lag, rolling, and calendar feature engineering for global forecasting setups.

- Deep learning for sequences: exposure to architectures such as DeepAR, N-BEATS/N-HiTS, Temporal Fusion Transformer, or LSTM/TCN-based models.

- Practical experience with forecasting libraries: statsmodels, Prophet, the Nixtla stack (statsforecast, mlforecast, neuralforecast), Darts, GluonTS, or sktime.

- Hierarchical or grouped forecasting and reconciliation; intermittent demand methods (Croston, ADIDA, and similar).

- Probabilistic forecasting: quantile regression, prediction intervals, and pinball loss.

Engineering and Production :

- Expert-level Python (pandas, NumPy, scikit-learn) and strong SQL.

- Experience deploying ML systems to production: pipelines and orchestration (Airflow, Dagster, or similar), containerization (Docker), CI/CD, and cloud platforms (AWS, GCP, or Azure).

- Working knowledge of distributed or large-scale data processing (Spark, Dask, or warehouse-native compute) for forecasting thousands to millions of series.

- MLOps fundamentals: experiment tracking (MLflow or W&B), model registry, retraining automation, and drift and performance monitoring.

Analytical and Business Acumen :

- Rigorous approach to time-based cross-validation, backtesting design, and avoiding data leakage.

- Ability to connect forecast accuracy to business outcomes (stockouts, overstock, service levels, working capital).

- Strong communication skills, with the ability to explain model choices, uncertainty, and trade-offs to non-technical stakeholders.

Good to Have :

- Domain experience in retail, e-commerce, CPG, logistics, energy, or quick-commerce demand planning.

- Exposure to causal inference or uplift modeling for promotion and price effect estimation.

- Experience with demand sensing using high-frequency signals or external data enrichment.

- Familiarity with optimization downstream of forecasting (inventory optimization, replenishment, S&OP integration).

- Contributions to open-source forecasting tools, or publications and blogs in the space.

Qualifications :

- Bachelor's or Master's degree in Computer Science, Statistics, Applied Mathematics, Operations Research, Economics, or a related quantitative field.

- 4 to 6 years of hands-on experience in data science or ML, with at least 2 to 3 years focused on time-series or demand forecasting in production settings.