Geospatial decision support
Valencia Bike Equity
An exploratory, reproducible system for reviewing territorial equity in bicycle parking across Valencia.
Public data Valencia Open Data
- parking points processed
- 4,316
- MCDA weight scenarios
- 10,000
Data science portfolio · Universitat Politècnica de València
Lianghao Zhan
Reproducible public-data projects in data engineering, urban analytics, forecasting, NLP retrieval and machine learning.
Data Science student seeking an internship in Valencia, hybrid or remote within Spain.
Working principle
Work that can be inspected, not just described.
Each project has a defined question, documented data source, executable pipeline, measured outputs and an explicit boundary around what the output does not prove.
These are portfolio studies over public data. They are not presented as work delivered for companies or as deployed production systems.
Selected work
Six projects built around data decisions.
Filter by domain or search the methods and artifacts used in each case study.
6 case studies visible
Operations research and mobility
Valenbisi Pulse
A reproducible snapshot control center for station risk, local pressure, minimum-cost rebalancing and conservative stress tests.
Public data CityBikes API / versioned local sample
- stations in reproducible sample
- 30
- critical stations at base thresholds
- 21
Forecasting and data platform
Spain Electricity Demand Forecast Lab
A weather-informed daily demand forecasting platform with temporal backtests, calibrated intervals and versioned SQL analytics.
Public data REE API, Open-Meteo ERA5 and Spanish holidays
- daily demand observations
- 2,557
- expanding backtest windows
- 12
Governed supervised learning
E-commerce Conversion ML
A leakage-aware classification system with model comparison, calibration, cost policies, cohort stress tests, monitoring and an API.
Public data UCI Online Shoppers Purchasing Intention Dataset
- public e-commerce sessions
- 12,330
- deployment-safe ROC-AUC
- 0.7602
NLP retrieval and open data
Valencia Open Data Navigator
A hybrid retrieval system for a versioned Valencia CKAN catalog, with benchmarked ranking, API and an interactive discovery interface.
Public data Valencia Open Data CKAN catalog
- datasets in versioned snapshot
- 296
- manual relevance queries
- 20
Data engineering and SQL
Valencia Air Quality Lakehouse
A reproducible DuckDB warehouse for Valencia public air-quality data, with source contracts, observability controls, SQL marts, API and dashboard.
Public data Valencia Open Data hourly air-quality distribution
- station-hour rows in snapshot
- 449k
- long-format measurements
- 6.04M
No project matches these filters.
Portfolio index
A compact evidence ledger.
The same filters above update this table. Values are pulled from committed project runs and link back to the corresponding case study or repository.
| Case study | Decision lens | Measured evidence | Reviewable artifacts | Review |
|---|---|---|---|---|
| Valencia Bike EquityGeospatial decision support | Geospatial · Urban analytics | 10,000MCDA weight scenarios | Offline snapshot · Spatial diagnostics · MCDA simulation | Repository |
| Valenbisi PulseOperations research and mobility | Operations research · Mobility | 21critical stations at base thresholds | Data validation · Risk scoring · Linear program | Repository |
| Spain Electricity Demand Forecast LabForecasting and data platform | Forecasting · Data engineering | 12expanding backtest windows | ETL manifests · SQL marts · Temporal backtests | Repository |
| E-commerce Conversion MLGoverned supervised learning | Machine learning · MLOps | 0.7602deployment-safe ROC-AUC | Model pipelines · Ablation study · Calibration | Repository |
| Valencia Open Data NavigatorNLP retrieval and open data | NLP · Information retrieval | 20manual relevance queries | Catalog snapshot · BM25 and LSA · RRF and MMR | Repository |
| Valencia Air Quality LakehouseData engineering and SQL | Data engineering · SQL | 6.04Mlong-format measurements | Source contract · Dimensional model · SQL quality marts | Repository |
Evidence ledger
What is visible in the repositories.
Models are not only compared. Their inputs, validation design, uncertainty, operational assumptions and failure modes are surfaced as first-class artifacts.
01
Provenance
Public sources, versioned snapshots and data cards make the origin of each dataset inspectable.
02
Validation
Temporal splits, leakage checks, quality controls and statistical comparisons are attached to the decision context.
03
Reproducibility
Source code, tests, linting, generated reports and CI workflows are available in every project repository.
Technical practice
A portfolio built around decisions.
The tools change with the problem; the working habits stay constant.
01
Data foundations
Python, pandas, SQL, DuckDB, validation checks, manifests and carefully scoped data transformations.
02
Modeling and inference
scikit-learn pipelines, calibrated probabilities, temporal backtesting, bootstrap intervals, spatial diagnostics and optimization.
03
Communication and quality
Plotly and Streamlit exploration, model and decision cards, pytest, Ruff, Docker where useful and GitHub Actions.
Availability
Looking for a Data Science internship.
Open to roles in data analysis, data science, machine learning, business intelligence, automation and junior data engineering in Valencia, hybrid or remote in Spain.