ML Lead

Role Summary
We are looking for a Machine Learning Engineer with strong expertise in predictive modelling and forecasting to design, build, train, and deploy production-grade ML systems. You will own the full model lifecycle — from data exploration and feature engineering to training, evaluation, deployment, and monitoring — and partner with engineering and product teams to ship predictive and forecasting capabilities that deliver measurable business outcomes.
Key Responsibilities
    •    Perform exploratory data analysis and feature engineering on large, real-world datasets, including temporal, transactional, and hierarchical data.
    •    Build, train, and tune predictive models and forecasting solutions using classical statistical methods, machine learning, and deep-learning techniques.
    •    Develop time-series forecasting models (short-term, mid-term, long-term horizons) with appropriate handling of seasonality, trend, intermittency, and external regressors.
    •    Design multi-level / hierarchical forecasts (e.g., by product, location, channel, time bucket) and reconcile them across levels.
    •    Run iterative training–validation–retraining cycles; track parameter tuning, feature importance, model drift, and prediction confidence intervals.
    •    Benchmark multiple model families (e.g., ARIMA/SARIMAX, exponential smoothing, Prophet, gradient boosting, deep-learning forecasters) and justify selection against agreed accuracy and business KPIs.
    •    Quantify uncertainty in predictions (prediction intervals, probabilistic forecasts) so downstream consumers can make risk-aware decisions.
    •    Package models as services or pipelines and collaborate with engineering to integrate predictions into production applications.
    •    Establish operational feedback loops using actuals-vs-forecast variance, real-world outcomes, and user signals to continuously improve models.
    •    Apply explainable-AI techniques so predictions can be reviewed, trusted, and overridden where needed.
    •    Document model design, assumptions, evaluation metrics, and limitations for stakeholder review.
Must-Have Skills
    •    7+ years building and deploying production ML models, with 2+ years on prediction and forecasting problems.
    •    Strong Python: pandas, NumPy, scikit-learn, statsmodels, Prophet, XGBoost/LightGBM, PyTorch or TensorFlow.
    •    Deep working knowledge of time-series and forecasting techniques:
    •    Classical: ARIMA/SARIMAX, ETS / Holt-Winters, state-space models.
    •    ML-based: gradient boosting on temporal features, lag/rolling features, regression with exogenous variables.
    •    Deep-learning forecasters: LSTM/GRU, N-BEATS, NHITS, Temporal Fusion Transformer (a plus).
    •    Solid understanding of forecast accuracy metrics (MAPE, sMAPE, WAPE, RMSE, MASE, pinball loss) and proper backtesting / walk-forward validation techniques.
    •    Hands-on with MLOps fundamentals: experiment tracking (MLflow / Weights & Biases), model versioning, reproducible pipelines, CI/CD for ML.
    •    Comfortable working with SQL and modern data platforms (Snowflake, Databricks, BigQuery, or equivalent).
    •    Strong fundamentals in statistics, evaluation methodology, and bias/variance trade-offs.
Nice to Have
    •    Experience with demand forecasting, sales forecasting, inventory prediction, or capacity planning use cases.
    •    Familiarity with hierarchical forecast reconciliation (top-down, bottom-up, MinT) and intermittent-demand methods (Croston, TSB, ADIDA).
    •    Experience deploying models on cloud platforms (AWS, Azure, GCP) and integrating them with operational systems.
    •    Familiarity with feature stores, streaming data, and real-time / batch inference patterns.
    •    Knowledge of explainable AI techniques (SHAP, LIME) applied to forecasts.
    •    Exposure to containerization (Docker, Kubernetes) and orchestration tools (Airflow, Kubeflow).