Title : Data Scientist (Contract) HR Analytics (Attrition, Survival Analysis, Causal Inference)

Engagement type : Contract (part-time/full-time), remote/hybrid (India)

Duration : 8- 16 weeks (extendable)

Start : ASAP

Role overview :

Were looking for a hands-on Data Scientist contractor to build and productionize attrition risk and retention insights for HR stakeholders. Ideal candidates have shipped attrition models in real systems and are comfortable with survival analysis and causal inference to drive decision-making.

Responsibilities :

- Build attrition prediction + time-to-exit models using survival analysis (KaplanMeier, Cox PH; bonus: AFT, competing risks).

- Design and implement causal inference for HR interventions (Propensity Score methods, RDD; bonus: DiD, causal forests/uplift).

- Translate HR questions into measurable outcomes (time-to-attrition, hazard ratios, retention lift, intervention ROI).

- Productionize pipelines : feature generation, model training, monitoring, and periodic retraining.

- Partner with HR / People Analytics teams to build explainable outputs and decision playbooks.

- Ensure privacy-safe modeling (PII handling, fairness/bias checks).

Must-have skills :

- 4- 8+ years in DS/Applied ML with at least one production deployment.

- Strong experience in survival analysis (Cox PH assumptions/diagnostics, censoring, time-varying covariates is a plus).

- Strong experience in causal inference (PSM/IPW, RDD; understanding of identification assumptions). Python stack : pandas/numpy/sklearn; stats libraries (statsmodels/lifelines); SQL.

- Ability to communicate results to business stakeholders.

Nice-to-have :

- HR domain experience : attrition, engagement, comp/benefits, performance, hiring.

- MLOps : MLflow, Airflow, Docker, CI/CD, model monitoring.

- Worked with large-scale data (Spark/Databricks).

Deliverables (what good looks like) :

- Survival-based attrition model + documented assumptions and diagnostics

- Causal analysis of 1- 2 interventions with clear methodology + effect sizes

- Deployment-ready code, monitoring plan, and stakeholder-ready insights