Location: Toronto, Canada
Email: priya.raghunathan@example.com
Phone: +1 416 555 0142
GitHub: github.com/example
Data Scientist with 6 years turning messy operational data into decisions people actually act on. Strongest where modelling meets deployment: I ship models into production and stay responsible for them afterwards. Looking for a team that measures data science by outcomes rather than notebooks.
Retention spend was spread evenly across all at-risk accounts because nobody could rank them. Built a gradient-boosted churn model on 18 months of billing and support data, calibrated so the score could be read as a probability rather than a rank. Shipped it behind an API the CRM calls nightly. Targeted campaigns on the top decile cut churn in that group by 19% against a held-out control, on ~40k monthly accounts.
Demand forecasts were evaluated by eyeballing charts, so nobody could say whether a change helped. Built a backtesting harness that replays any candidate model over historical windows and reports error by horizon and by segment. It caught a seasonal regression that had been live for two quarters, and made model changes an argument about numbers rather than opinion.
Data Scientist | Apr 2021 - Present
- Own the churn and demand-forecasting models end to end: Python, scikit-learn, XGBoost, Airflow.
- Serve models through a FastAPI service on Kubernetes, with drift monitoring and weekly retraining.
- Work directly with the retention team to design experiments and read the results honestly.
- Cut the forecasting pipeline runtime from 6 hours to 40 minutes by rewriting joins in DuckDB.
Data Analyst, then Data Scientist | Jul 2018 - Mar 2021
- Built the reporting layer in dbt and BigQuery that replaced a spreadsheet-based weekly pack.
- Developed a route-anomaly detector that flagged ~200 suspect trips/month for review.
- Ran the A/B analysis for pricing changes across 12 regional markets.
- Languages: Python, SQL, R.
- Libraries: scikit-learn, XGBoost, statsmodels, pandas, Polars.
- Methods: Causal inference and A/B testing, time-series forecasting, calibration, survival analysis.
- Pipelines: Airflow, dbt, BigQuery, DuckDB, Spark.
- Deployment: FastAPI, Docker, Kubernetes, MLflow, drift monitoring.
- Visualisation: Plotly, Streamlit, Looker.
MMath Statistics | 2016 - 2018
BSc Mathematics | 2013 - 2016