Real Estate Fair Price Estimator
A Big Data pipeline: is this property fairly priced?
System architecture
- 01
Data Lake
Historical property transactions, from raw to served
- 02
Orchestration
Apache Airflow drives the processing
- 03
Transformation
Distributed Spark (PySpark), columnar Parquet storage
- 04
Search
Elasticsearch indexing + Kibana dashboards
- 05
Verdict
FastAPI — price range, confidence, rental yield
You give the system a property (city, type, surface, price) and it compares it against thousands of historical transactions to deliver a verdict — underpriced, fairly priced or overpriced — with an estimated price range, a confidence level and the rental yield. Orchestration is handled by Airflow, transformation by Spark (PySpark) with Parquet storage, search by Elasticsearch, and visualisation by Kibana plus a custom FastAPI UI.
What was built
- Clean Data Lake architecture, from raw to served
- Apache Airflow orchestration of the processing
- Distributed Spark (PySpark) transformation, columnar Parquet storage
- Elasticsearch indexing + Kibana dashboards
- FastAPI UI returning a price verdict + range + rental yield
- Live demo: ismyhouseexpensive.netlify.app
The real challenge
Beyond the model, the real work was the data architecture: moving data cleanly from a raw lake to a real-time served answer. That is exactly the chain of a Data Engineer role.
Next project
TeamUp