Skip to content
View dtortosa's full-sized avatar

Highlights

  • Pro

Block or report dtortosa

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
dtortosa/README.md

Diego Salazar Tortosa PhD | Data & ML Scientist

I build machine learning models and the data pipelines that feed them. PhD in Bioinformatics and 11+ years as a Data & ML Scientist across Europe and the US. Extensive experience with large, high-dimensional datasets in multiple fields. I help teams turn messy data into validated insights.

🛠️ Tech Stack

  • Machine Learning & Statistics: Tabular foundation models (pretrained transformers, e.g., TabPFN), Deep Neural Networks (multilayer perceptrons, Keras), Tree-Based Ensembles (XGBoost, Random Forest), GLMs/GAMs, Elastic Net, Bayesian Statistics, Nested Cross-Validation, Bayesian Optimization, Explainable AI
  • Scientific Computing: Python, R, NumPy, Pandas, PySpark, Bash, AWK, Parallel and Distributed Processing
  • Infrastructure: Slurm, Nextflow, Docker & Apptainer, Azure, PostgreSQL, Git/GitHub Actions
  • LLM Applications: RAG, FastAPI, Celery, pgvector, PydanticAI, LangChain, LangGraph, Langfuse, LLM evals

📈 Featured Projects

  • Human genome ML framework: nested cross-validation and Bayesian optimization across multiple model classes on ~180 GB of compressed genetic data (~1 billion rows; PySpark, multiprocessing, Slurm); explainable AI to interpret deep neural networks.
  • Global pine distributions under climate change: ensemble species distribution models with 100k+ GCM × RCP projections, methodological innovations improving the state of the art, validated on spatially independent data. Published in Diversity and Distributions (2024).
  • Email triage LLM workflow: event-driven pipeline with deterministic routing, RAG over pgvector, LLM-as-judge evaluation and Langfuse tracing, deployed with Docker on Hetzner.

Pinned Loading

  1. genome-wide-analysis-trainability-australia genome-wide-analysis-trainability-australia Public

    Genomic pipeline to predict physical performance from genome-wide data: QC, imputation to 3.2M variants (1,204 individuals), and parallelized polygenic scoring (PySpark, Slurm) validated over 100 t…

    Python

  2. novel-ml-method-climate-change-prediction novel-ml-method-climate-change-prediction Public

    Ensemble species distribution modeling (RF, GAM, GLM) predicting climate change impacts on global pine distributions by 2070: 100k+ projections across GCMs and RCPs, validated on spatially independ…

    R

  3. twitter-forex-ml-pipeline twitter-forex-ml-pipeline Public

    Tests whether central bank sentiment from 3.6M tweets (ECB, Fed) improves EUR/USD forecasts over a random-walk baseline, using time-series cross-validation.

    Jupyter Notebook

  4. ai-business-intelligence-assistant ai-business-intelligence-assistant Public

    Purdue GenAI capstone: LangChain agent that chooses between sales data analysis (Pandas), document retrieval (FAISS RAG) and Wikipedia search, with answer evaluation in a Streamlit dashboard.

    Jupyter Notebook

  5. hr-knowledge-rag-agent hr-knowledge-rag-agent Public

    RAG prototype for question answering over a corporate HR handbook: PDF chunking, Chroma vector search, and conditional Jinja2 prompt templates, with a Gradio UI.

    Jupyter Notebook