From processing cosmic ray events at the HAWC Observatory to rebuilding agricultural credit analysis at a national financial institution — twenty years of turning complex, untamed data into decisions that move organizations forward. The source and format change every time. The discipline does not.
"Translate data into actionable insights through data storytelling, empowering strategic decisions." — Data Scientist · Role Description
That is what I have done across seven industries, from particle physics to agricultural finance. The foundation came from the HAWC Observatory — a multinational cosmic ray experiment with distributed computing, petabyte-scale event data, and statistical analysis across institutions in the US and Mexico. That standard held: when the data is hard, the discipline matters most.
In industry, the data changed. At D3Clarity, I turned thousands of unstructured public school board meeting minutes into a competitive intelligence dashboard used daily by a sales team. The core task — classify text, surface patterns, structure the unlabeled — was familiar ground: my MSc thesis at the National Institute of Astrophysics addressed automatic text classification through prototype-based learning. That was 2009. At FIRA, I replaced manual Excel workflows with a governed pipeline integrating satellite imagery, climate models, and regulatory databases into credit decisions for agricultural producers. At CITEIM, I applied Transformer architectures to sign language recognition. The tools advanced. The method held.
Every engagement followed the same sequence: establish data quality, build outputs accessible to non-technical users, demonstrate value early, then scale. What makes work at this level engaging is the specific challenge it carries: shaping how data science functions inside an organization — building the culture, not executing within one that already exists. My technical trajectory has followed the broader AI evolution, from classical ML through production systems to generative AI. That progression is something to build on, not replace.
Excerpts from the role requirements, with context from the work.
"Collaborate within Agile squads to design and implement machine learning models that solve real business problems."
Four years at D3Clarity in Agile/Scrum with Jira across distributed international teams. End-to-end ML delivery — NLP classification from unstructured public documents to Power BI dashboards used by sales teams. Every project: PoC → production.
"Manage the model lifecycle: from ideation and data preparation to deployment and scaling on Cloud platforms using modern MLOps practices."
At FIRA: public source ingestion → geospatial validation → climate risk integration → governance protocols → Apache Superset dashboards. Stack: Python, PostgreSQL, Linux, AWS EC2, Azure ML, SageMaker. AWS Cloud Practitioner certified. Google Workspace administrator (own domain, independently configured) — platform selection by fit, not familiarity.
"Translate data into actionable insights through data storytelling, empowering strategic decisions."
Dashboards built for non-technical credit analysts at FIRA — no training required for daily use. Executive KPI reporting at Punto Singular. Taught Tableau-based dashboarding to Business School students at Tec de Monterrey. Power BI · Tableau · Looker · Apache Superset.
"Contribute to building a strong data culture, ensuring quality, ethics, and compliance in every solution."
Defined governance frameworks and quality protocols at FIRA. Incorporated environmental compliance layers into credit assessment. Led a mentorship program building the next generation of practitioners. Semarchy xDM certified — MDM, Data Profiling, Data Lineage, Golden Records.
"Degree in a quantitative field. Fluency in English. Italian is a plus."
PhD Candidate in Computer Physics · MSc Computer Science (ML/NLP) · BSc Physics & Mathematics. English: professional — 4+ years with US and European clients. Italian: basic-intermediate — ongoing private tutoring, native Spanish provides structural advantage for rapid progression.
Scientific rigor applied across an expanding surface area of domains and industries.
Each one began with a focused hypothesis. Each scaled once value was demonstrated.
A client needed to sharpen their sales strategy for school district products across the US. School board meeting minutes are public record. I built the full pipeline: automated scraping → NLP processing → ML topic classification → Power BI dashboard mapping which products to pitch to each district based on what they are actively discussing at the board level. Public data transformed into a scalable commercial tool.
Replaced ad-hoc spreadsheet processes with an automated pipeline that integrates satellite-derived vegetation health indices, multi-source climate risk models, and regulatory databases (protected environmental zones). Credit analysts receive structured, repeatable reports — no manual data aggregation, no ad-hoc interpretation. Data governance protocols established across all source feeds.
Designed and implemented Transformer-based models for continuous sign language recognition, integrating cloud services and pre-trained language models to enable real-time interpretation. Brought current-generation AI architectures to a problem with direct human communication impact — demonstrating that modern ML methods generalize across domains when the fundamentals are solid.
Tools are means, not ends — selected for fit, not for familiarity.
Values, language, and context.
Available to relocate. Open to international positions. Logistical transition is not a constraint.
"The source and format of the data change every time; the discipline does not."