Data & AI Specialist | PhD | Applied AI, machine learning and open data for climate, maritime and sustainability decisions
I build data and AI tools that turn messy public data into answers people can act on: open datasets, machine-learning models, computer vision, and LLM tools (MCP) that cite their sources. Most of my work is on ports, shipping and climate. Open to data, applied AI and sustainability analytics roles in the Netherlands, across Europe and remotely.
| Competency | Project | Result | Links |
|---|---|---|---|
| AI engineering (MCP) | port-emissions-mcp | MCP server with six tools that lets Claude answer questions on 247 Brazilian port installations and estimate at-berth CO₂, citing source and method in every answer; offline test suite | Code |
| Open data / data engineering | antaq-port-statistics | Annual cargo, berthings and port times for 247 installations, 2010-2026, aggregated so national totals match the regulator's official figures | Code · DOI · Live |
| Computer vision | BluePort AI | Waste photos sorted into six recycling streams with frozen CLIP + linear probe: 94.6% cross-validated accuracy on 11,451 images, runs in the browser | Try it · Code · HF Space |
| Machine learning | vessel-efficiency-ml | Leakage-free scikit-learn pipeline predicting ships' CO₂ efficiency grades (A-E) from EU MRV data | Code · DOI |
| Regulatory analytics | shipping-methane-monitor | Verified CH₄ and N₂O from 16,664 ships (EU MRV 2024-2025): the same ships cut CO₂ 5% and raised CH₄ 21% | Live · Code · DOI |
| Python library | maritimeco2 | Citable library for OPS-avoidable at-berth CO₂ (IMO Fourth GHG Study method), used by the MCP server and decarbport.com | Code · DOI |
| SQL | antaq-port-sql | DuckDB star-schema warehouse of Brazilian port operations with documented queries (CTEs, window functions) | Code · DOI |
| Decision support | SDG port attributes | Peer-reviewed assessment framework: 84 indicators, 20 dimensions, 8 SDGs (Marine Policy 2025) | Live · Code · DOI |
More: shipping-carbon-costs (EU ETS cost of 3,303 companies) · eu-ports-no2 (Sentinel-5P) · Brazil_Vessel_Call_Intelligence (30,972 port calls) · textmining (PDFs to VOSviewer networks) · full list on ribeirocunha.com
real-world question → public data → cleaning, validation, harmonisation
→ analysis and models (statistics, ML, CV, NLP) → open dataset with DOI
→ tool people use (web app, library, MCP server for LLMs) → decision
Every project ships with documented methods, stated limitations, a licence and citation metadata.
Data I know well: ANTAQ (Brazil's waterway regulator), EU MRV / THETIS, EU ETS, Eurostat maritime, Copernicus Sentinel-5P, IMO GHG Study methods.
Full Professor at the Federal University of Maranhão (Brazil) · PhD in Accounting and Finance, Universidad de Zaragoza · Postdoctoral research, Erasmus University Rotterdam (2024-2025). Reviewer for 26 international journals.
- Cunha, D. R., & Pereira, N. N. (2027). Mapping ESG evolution in Brazilian ports: A text-mining approach to sustainability reporting. Marine Policy, 195, 107287. doi:10.1016/j.marpol.2026.107287
- Cunha, D. R., Costa, M. D. C., Oliveira, C. B. M., & Pereira, N. N. (2025). SDG attributes: A sustainability assessment framework for Brazilian ports. Marine Policy, 181, 106841. doi:10.1016/j.marpol.2025.106841
Retrieval-augmented generation over port and sustainability documents · evaluation of LLM tool use · agents that combine MCP servers with open datasets.