Zrive DS Course
6 end-to-end data science modules.
End-to-End Pipeline
Comprehensive implementation covering data acquisition, processing, modeling, and production deployment.
Stack Insights
- Primary LanguagePython 3.10
- LibrariesScikit-learn, Pandas
- DeploymentDocker, FastAPI
“Built with a focus on production-ready code and rigorous statistical evaluation.”
Program Brochure
Official Zrive Applied Data Science (Z-DS) program guide, a 15-week curriculum, modules, instructors and real-company project.
account_treeCurriculum Modules
Climate Data Analysis
Open-Meteo API integration to pull hourly weather data for Madrid, London, and Rio de Janeiro (2010 to 2020). Built a resilient HTTP client with exponential backoff, then used pandas for aggregation and matplotlib for multi-panel visualisations.
E-commerce EDA
Exploratory data analysis on a grocery e-commerce dataset (S3/Parquet). Five datasets: orders, regulars, abandoned carts, inventory, and users. UK NUTS1 regional analysis, user segmentation, product prevalence, and price distributions.
Push Notification Propensity (Ridge)
Binary classification model to predict push notification engagement. Ridge Logistic Regression with time-aware train/validation split by order date, grid search over C, optimised for PR AUC. Model serialised with joblib.
Push Notification Propensity (XGBoost)
Same propensity problem upgraded to XGBoost with a production-grade event-driven pipeline. PushModel class, handler_fit and handler_predict handlers, shared utils.py, and a Jupyter notebook for exploration.
Financial Report Scoring (LightGBM)
LightGBM model trained on quarterly financial reports. Binary target: does this stock outperform the S&P 500 over the next year? Rolling time-series splits, model diagnosis loop, and weighted portfolio return as the evaluation metric.
Basket Recommendation API (FastAPI)
Basket recommendation model served as a FastAPI REST API. GET /status, POST /predict, GET /metrics endpoints. Full routers/services/utils/models/exceptions architecture, Pydantic schemas, and load testing with attack.sh and report.sh.
Business Translation
Frameworks for translating data science outputs into clear business communication. Covers structuring recommendations, quantifying impact, and presenting model results to non-technical stakeholders.