DataHut-DuckHouse
ActiveA multi-tenant analytics platform, local or cloud: Arrow Flight ingestion, hybrid DuckDB and Iceberg storage (MinIO/S3), dbt transformations and Trino queries.
- Python
- DuckDB
- Apache Iceberg
- Arrow Flight
- dbt
- Trino
- MinIO
Data & Analytics
ETL pipelines, lakehouse architectures and dashboards, with steady attention to data quality.
Experience
Data Analyst · Solutions S.A.
- Analytics architecture on Microsoft Fabric (Lakehouse, Data Warehouse).
- Power BI dashboards and KPI tracking; helped raise the share of client sites above 95% performance from about 20% to about 60%.
- ETL pipelines and healthcare data quality, including machine-learning-based deduplication.
- Mapping with QGIS and Power BI.
- SQL Server, MySQL and PostgreSQL databases.
Data Engineer (volunteer) · AYITISTATS
- Unifying statistical data for a Google Data Commons instance on Google Cloud (GCP).
Lecturer, Mathematics for Computer Science · Université Quisqueya
- Mathematics for computer science, taught with a focus on its uses in data science, software engineering and web technologies.
Teaching Assistant, Data Science & AI · Université Quisqueya
- Supporting a hands-on curriculum run with Akademi and aligned with the Flatiron School immersive program.
- Data wrangling, exploratory analysis, model building and deployment with Python, Pandas, NumPy, Scikit-learn and Jupyter.
- Code reviews, Q&A sessions and contributions to curriculum refinement.
Projects
Nyansa
ActiveA self-hosted AI second brain on a private server: seven Docker Compose services (local LLMs with Ollama, Qdrant vector store, n8n workflows) wired to an Obsidian vault. Built on n8n's self-hosted AI starter kit.
- Docker Compose
- n8n
- Ollama
- Qdrant
- PostgreSQL
- Open WebUI
Scientific Assistant
ActiveA local multi-agent system: eight specialized agents (planning, symbolic mathematics, numerical methods, statistics, literature, review, writing, memory) and hybrid vector and keyword RAG. Produces executable Python code and citations.
- Python
- LangChain
- Ollama
- ChromaDB
- RAG
Simulation chlordécone
ActiveA non-autonomous compartmental ODE model of chlordecone contamination under seasonal rainfall forcing, with a Shiny app to simulate human exposure and intervention strategies. Code accompanying a preprint.
- R
- Shiny
- Non-autonomous ODEs
- renv
Medical Triage System
PrototypeA C++ triage application: a Random Forest model classifies patients into three levels (critical, moderate, low) from their symptoms and risk factors.
- C++
- Random Forest
- CSV
Freudz, analyse de sentiment
CompletedReview sentiment classification on a bag-of-words representation, with implementations of Perceptron, Average Perceptron and Pegasos, hyperparameter tuning and a Streamlit interface.
- Python
- NumPy
- Streamlit
Skills
Languages
- Python
- SQL
- R
Databases
- SQL Server
- PostgreSQL
- MySQL
- DuckDB
Data engineering
- Microsoft Fabric (Lakehouse, Data Warehouse)
- ETL
- dbt
- Apache Iceberg
- Trino
- Arrow Flight
- Spark SQL
- Google Cloud (GCP)
BI & visualization
- Power BI
- QGIS
- Shiny
Machine learning & AI
- Scikit-learn
- Pandas
- NumPy
- RAG
- Multi-agent systems
- Local LLMs (Ollama)
DevOps & tooling
- Docker
- Linux
- Git
Mathematics & scientific computing
- Statistics & probability
- Compartmental modelling
Education
Master's in Mathematics and Applications, MOAD track
Mathematical modelling and decision support tools, environment and health focus
MicroMasters in Statistics and Data Science
Methods track
Bachelor's degree in Mathematics
Civil Engineering degree
Certifications
Microsoft Certified: Fabric Data Engineer Associate (DP-700)
Microsoft Certified: Fabric Analytics Engineer Associate (DP-600)
Apache Spark SQL for Data Analysts
Google Data Analytics Professional Certificate
Google Business Intelligence Professional Certificate