We are seeking a skilled Time Series Forecasting Engineer to enhance existing Python microservices into a modular, scalable forecasting engine. The ideal candidate will have a strong statistical background, expertise in handling multi-seasonal and intermittent data, and a passion for model interpretability and real-time insights.
Key Responsibilities
Develop and integrate advanced time-series models: MSTL, Croston, TSB, Box-Cox.
Implement rolling-origin cross-validation and hyperparameter tuning.
Blend models such as ARIMA, Prophet, and XGBoost for improved accuracy.
Generate SHAP-based driver insights and deliver them to a React dashboard via GraphQL.
Monitor forecast performance with Prometheus and Grafana; trigger alerts based on degradation.
Core Technical Skills
Languages
: Python (pandas, statsmodels, scikit-learn)
Time Series
: ARIMA, MSTL, Croston, Prophet, TSB
Tools
: Docker, REST API, GraphQL, Git-flow, Unit Testing
Database
: PostgreSQL
Monitoring
: Prometheus, Grafana
Nice-to-Have
: MLflow, ONNX, TensorFlow Probability
Soft Skills
Strong communication and collaboration skills
Ability to explain statistical models in layman terms
Proactive problem-solving attitude
Comfort working cross-functionally in iterative development environments
Job Type: Full-time
Pay: ?400,000.00 - ?800,000.00 per year
Application Question(s):
Do you have at least 2 years of hands-on experience in Python-based time series forecasting?
Have you worked in retail or manufacturing domains where demand forecasting was a core responsibility?
Are you currently authorized to work in India without sponsorship?
Have you implemented or used ARIMA, Prophet, or MSTL in any of your projects?
Have you used Croston or TSB models for forecasting intermittent demand?
Are you familiar with SHAP for model interpretability?
Have you containerized a forecasting pipeline using Docker and exposed it through a REST or GraphQL API?
Have you used Prometheus and Grafana to monitor model performance in production?
Work Location: In person