I'm Diego, and I've been turning complex data into validated insights for over 11 years
-
Do you have data piling up but no clear answer to the questions that matter?
-
Does your model look good in testing, but you are not sure it will hold on new data?
-
Do you need to know which variables actually drive an outcome, not just a prediction?
-
Looking for PhD-level rigor without building a full data science team?
About me
I build machine learning models and the data pipelines that feed them, for teams that need predictive models validated rigorously enough to act on.
I have over 11 years of experience as a Data & ML Scientist across Europe and the US, and a PhD in Bioinformatics. My work spans the modeling of the human genome at the University of Arizona, forecasting species distributions under climate change in collaboration with Switzerland's WSL, and cancer companion diagnostics at expressTEC. The method is consistent across domains: scalable data pipelines, systematic comparison of algorithms, nested cross-validation, validation on independent data, and explainability techniques so results can be interpreted by domain experts. I also follow recent advances in ML for tabular data, including tabular foundation models.
When a problem calls for LLMs, I build those applications too (RAG, workflow automation with FastAPI, tracing and evals with Langfuse), using the same evaluation discipline.
🛠️ Tech Stack: Python, R, SQL, NumPy, Pandas, scikit-learn, Tabular foundation models (pretrained transformers, e.g., TabPFN), Deep Neural Networks (multilayer perceptrons, Keras), XGBoost, Random Forest, Elastic Net, Bayesian Optimization, Nested Cross-Validation, Explainable AI, PySpark, Slurm, Nextflow, Docker/Apptainer, Azure, GitHub Actions, Git | LLM: RAG, FastAPI, Celery, pgvector, PydanticAI, LangChain, LangGraph, Langfuse, LLM evals
Why work with me?
Here's what sets me apart, and how I can help your team:
-
Scientific Rigor
With a PhD in Bioinformatics and over 11 years as a Data & ML Scientist, I treat every model as a hypothesis to test. Nested cross-validation, validation on independent data and honest baselines tell you how a model will perform on data it has never seen, not just on the data it was trained on.
-
Models You Can Interpret
A prediction is only useful if you can trust and explain it. I use explainable AI techniques (permutation importance, accumulated local effects) to show which variables drive each result, so domain experts can check the model against what they know and act on it.
-
Large, Complex Data
I have built pipelines for ~180 GB of compressed genomic data, billions of genotypes and hundreds of thousands of climate projections, using parallel and distributed computing (PySpark, multiprocessing, Slurm). Messy, high-dimensional data is my starting point, not an obstacle.
-
The Right Model for the Problem
I do not start from a favorite algorithm. I benchmark candidates, from regularized linear models to gradient boosting, deep neural networks and tabular foundation models, and recommend the simplest one that performs well. When a problem calls for LLMs, I apply the same evaluation discipline.
What they say about my work
-
David J Bishop
Research Leader at iHeS (VU - Australia)
"I had the pleasure of working with Diego on a large-scale genomic study investigating genetic predictors of physical performance in military recruits. Diego led the analytical side of the project, processing and modelling data from over 3 million genetic variants across more than 1200 participants. His work involved rigorous quality control, population structure analysis, and the development of polygenic scores.
Diego’s technical and analytical skills are exceptional. He independently built and validated complex pipelines and implemented robust statistical models. His ability to translate raw data into meaningful insights was critical to the success of the project. He also communicated findings clearly to a multidisciplinary team, bridging the gap between computational biology and applied military research. Beyond his technical expertise, Diego is proactive, thoughtful, and collaborative. He consistently sought feedback, proposed improvements, and showed a strong sense of ownership. His contributions were instrumental in shaping the direction of the study.
I highly recommend Diego for any data science role that demands analytical rigor, scientific curiosity, and a collaborative mindset. He would be a valuable asset to any team working with complex datasets and looking to extract actionable insights.
Beyond his data analysis skills, he is also a great guy and easy to work with."
-
Jonatan R Ruiz
Director at iMUDS (UGR - Spain)
"I have worked with Diego over the last nine years across multiple projects. He has consistently led the processing and analysis of massive datasets, encompassing tens of millions of genetic variants across thousands of participants.
While I could praise Diego for many reasons, what stands out most is the rigor and neatness of his work. It is striking that a person who self-taught advanced programming and machine learning during his PhD reached such high standards in data processing, statistical analysis, and model interpretability. In every project, Diego took a proactive role in technical strategy, from data quality control and processing to selecting the machine learning architectures best suited for our needs.
Regardless of the challenges we encountered, Diego always identified the right path forward to achieve our project goals. I highly recommend him for any lead role in Data Science, Machine Learning, or analytical leadership. He deeply investigates the best approach for the project at hand and delivers results at the highest industry standards.
Happy to continue! show must go on!"
Frequently asked questions
How quickly can you start working on my project?
I can typically begin new projects within 1-2 weeks of contract signing. For urgent matters, I keep some flexibility and can potentially start sooner — just let me know your timeline during our initial consultation.
Do you require a minimum project size or commitment?
While I can accommodate projects of any size, I find that engagements of at least 10 hours allow for meaningful impact. This gives us enough time to understand your data, implement solutions, and deliver results you can act on. We can start with a small pilot project to ensure we're a good fit.
What industries do you have experience in?
I’ve delivered projects across human genomics, clinical diagnostics, exercise and health research, and environmental and spatial modeling. I am strongest where data is large, noisy or high-dimensional, and where a project needs predictive models that are also explainable. The methods themselves (data pipelines, model benchmarking, validation and interpretation) apply across sectors.
How do you handle data security and confidentiality?
I sign NDAs before starting any project and only access the data the project needs. When data cannot leave your organization, I can work within your existing infrastructure and security policies.
How do you communicate progress and results?
I share regular progress updates and hold check-in meetings at a frequency that suits the project. You receive documented code and clear reports of the analyses, results and recommendations, written so that non-specialists can act on them.
-
Let's have a virtual coffee together!
Want to see if we are a match? Let's have a chat and find out. Schedule a free 30-minute session to discuss your data challenges and explore how we can work together.