Data Scientist - Supply Chain & Inventory Analytics

Experience : 4 to 6 years

Location : Chennai

Note : We are looking for candidates who are currently serving notice or candidates who can join us immediately.

Key Responsibilities :

- Translate ambiguous operational problems stated by planners and engineers, not by data teams, into well-posed modelling problems with clear success measures.

- Build, validate, and productionise forecasting, optimisation, and regression models on real enterprise data (SAP MM / PM extracts, consumption history, purchase-order history, master data).

- Do the unglamorous data work properly : profiling, reconciliation, deduplication of SKU masters, handling intermittent and lumpy demand, and dealing with sparse or missing history.

- Design fallback and cold-start strategies so that models degrade gracefully rather than producing confident nonsense on thin data.

- Build explainability into every output - a planner must be able to see why a number moved before they will act on it.

- Partner with product, engineering, and design to ship models into a live application, including the schema contracts, validation rules, and retraining behaviour the application depends on.

- Run model monitoring and periodic retraining; investigate drift and degradation against ground truth from the field.

- Present findings and recommendations to senior client stakeholders - plant heads, materials management, and procurement leadership - in operational language, with the assumptions and limitations stated plainly.

- Document methodology to a standard that survives audit, client scrutiny, and your own handover.

Required Qualifications :

- Experience : 4 to 6 years in a data science, applied machine learning, or quantitative analytics role, with at least two years working on problems that reached production or live business use.

- Education : Bachelor's or Master's in Statistics, Mathematics, Computer Science, Operations Research, Industrial Engineering, Economics, or a related quantitative discipline.

- Programming : Strong Python - pandas, NumPy, scikit-learn, statsmodels. Comfortable writing clean, testable, reviewable code rather than notebook-only exploration.

- SQL : Confident with complex joins, window functions, and query performance on large operational tables.

- Time-series forecasting : Practical experience with classical and modern approaches - ARIMA/SARIMA, exponential smoothing, Prophet, gradient-boosted trees for tabular time series - and the judgement to know when a simple baseline is the right answer.

- Supervised learning : Solid grounding in regression and tree-based ensembles (XGBoost, LightGBM, Random Forest), including feature engineering, regularisation, cross-validation design, and honest error analysis.

- Statistical fluency : Distributions, uncertainty quantification, confidence and prediction intervals, hypothesis testing, and the ability to explain what a model does not know.

- Communication : Able to explain a model to a plant engineer with no statistics background and to defend it to a technically sharp reviewer, in the same week.

Preferred / Good to Have :

- Domain exposure to supply chain, inventory optimisation, spare-parts planning, MRO, maintenance planning, or procurement analytics.

- Familiarity with inventory theory - safety stock formulations, service-level targets, EOQ, reorder point logic, multi-echelon inventory concepts.

- Experience with intermittent and lumpy demand methods (Croston, SBA, TSB) - highly relevant for spare parts, where most SKUs move rarely.

- Optimisation experience : linear/mixed-integer programming with PuLP, OR-Tools, Gurobi, or similar.

- Working knowledge of SAP data structures (MM, PM, MRP) or comparable ERP extracts.

- Exposure to asset-heavy sectors - power generation, oil and gas, mining, heavy manufacturing, utilities, or process industries.

- Condition-monitoring, predictive maintenance, IoT sensor data, or reliability engineering (RCM, FMEA) exposure.

- MLOps practice : MLflow, Docker, CI/CD for models, experiment tracking, model registries.

- Cloud platforms - Azure, AWS, or GCP - and their data and ML services.

- Visualisation and storytelling : Power BI, Plotly, Streamlit, or building analytical front-ends that non-analysts actually use.

- Awareness of data-residency and security expectations in the Indian public-sector and regulated-enterprise context (single-tenant deployments, CERT-In alignment).