Role Summary

Transform GrepEye’s current Python microservices from using a single time-series forecasting model (Auto-ARIMA/SARIMAX) into a modular, scalable forecasting engine that:

• Handles multi-seasonality.
• Accounts for intermittent demand.
• Delivers driver-based insights in plain language.

Key Responsibilities

Model Development & Prototyping

• Implement advanced models: MSTL, log/Box-Cox transformations, Croston/T-SBOS for intermittent data.
• Enhance performance via rolling-origin cross-validation and hyperparameter tuning.

Model Blending

• Combine models (e.g., ARIMA + boosted trees/Prophet) to improve accuracy where needed.

Insights & Visualization

• Compute SHAP values for model interpretability.
• Send driver insight cards to a React dashboard via GraphQL.

Monitoring & Alerting

• Use Prometheus/Grafana to track forecast errors (MAPE/CI width) and alert on degradation.

Required Technical Skills

• Programming: Python (with pandas, statsmodels, scikit-learn).
• Time Series Expertise: Strong statistical background.
• Dev Tools: Docker, REST APIs, JSON, Git-flow.
• Data Handling: Basic PostgreSQL.
• Testing & Optimization: Unit testing, performance profiling of pipelines.

 Nice-to-Have Skills

• Tools: MLflow, ONNX.
• Models/Libraries: XGBoost, Prophet, TensorFlow Probability.
• Domain Experience: Demand forecasting in retail/manufacturing.

Soft Skills

• Ability to explain statistical ideas to non-technical audiences.
• Proactive in identifying data and modeling issues.
• Comfortable working collaboratively and iteratively with cross-functional teams.

 Ideal Candidate Profile

• A data-savvy time series engineer with:
• Proven experience in modular Python-based model development.
• Ability to blend statistical and machine learning models.
• Strong focus on model interpretability and real-time monitoring.
• Excellent communication and collaboration skills.