Cancer traceability and active case-management model
A process redesign without a single software licence, which took a public hospital into the national Top 5 for cancer management.
- Lean Six Sigma
- BPMN
- Value-Based Healthcare
- Gestión de casos
Chilean Ministry of Health · Office of Intelligence and Strategic Health Management
Data engineering, advanced analytics and artificial intelligence for public health.
From the clinical problem to the architecture, the model and the system in production.
I solve complex problems in public health systems with data engineering, advanced analytics and artificial intelligence: the ones conventional analysis cannot reach because of volume, because the clinical record is ambiguous, or because the decision behind them tolerates no automatic error. Twenty years in the Chilean public system, from clinical practice to models now running at national scale on open technology.
In production
Cancer suspicion detection in surgical waiting lists
Classifies ~475,000 diagnoses per cut and routes suspicious and indeterminate cases to medical clinical audit. A fine-tuned language model adds the affected organ in ICD-O3 code.
- NLP
- Embeddings
- XGBoost
- Llama 3.1 · QLoRA
- Human-in-the-loop
Quality validation of the waiting list warehouse
Runs 34 automated tests over each monthly cut on the raw and the transformed layer, routes problem records back to the facility that produced them, and certifies what goes into official reports. What fails the tests does not ship.
- Pointblank
- Capa bronce / plata
- R
- DuckDB
Anomaly detection in the National Waiting List Registry
Monitors volume, time on list, variance between services and discharge justifications, then notifies local management teams.
- Detección de anomalías
- Aprendizaje no supervisado
Probabilistic identity deduplication
Resolves the same person registered under different identifiers across 2.1 million records, correcting overstated demand.
- Record linkage
- Fellegi-Sunter
- Splink
- DuckDB
Since September 2026
What I am working on now
Since 1 September 2026 I have been part of the new department within the Office of Intelligence and Strategic Health Management at the Chilean Ministry of Health. The office concentrates the technical work of a 45-strong multidisciplinary team around seven strategic modernisation programmes, with national health interoperability as the target for 2030.
- 01
National Digital Prescription
Standardising and fully digitising the prescription and dispensing flow for medicines.
- 02
Shared Clinical Record and Interoperability
A platform so each person’s clinical history follows them securely across the public and private care network.
- 03
Digital Transformation of Waiting Lists
Intensive use of analytics, flow optimisation and artificial intelligence to relieve demand for appointments and surgery.
- 04
New National Immunisation System
A unified digital registry with continuous traceability for the national vaccination plan.
- 05
Health Map
A geo-intelligence platform structuring the supply and capacity of the care network.
- 06
Medicine Traceability
Monitoring the supply, storage and dispensing chain for medicines across health facilities.
- 07
Citizen Digital Health Portal
A unified interface giving patients access to their medical information, appointments and administrative processes in the public system.
Areas of work
Four lines of work that hold each other up
None of them works alone: the clinic defines the problem, engineering makes it tractable, the State scales it, and training keeps it alive over time.
Data engineering without a cluster
Tens of millions of hospital records processed on ordinary desktop machines, with embedded OLAP engines and open columnar formats. No clusters, no proprietary licences.
DuckDB · Apache Parquet · R · Python
Advanced analytics born in clinical practice
Models that work are not designed from the abstraction of data, but from the operational failures of direct care. Twenty years of clinical practice decide what is worth predicting.
analítica clínica · oncología · listas de espera · GES
Artificial intelligence inside the State
NLP over clinical referrals, anomaly detection and ensemble classification, always with mandatory human oversight. The algorithm triages; it never dismisses on its own.
NLP · embeddings · GBDT · human-in-the-loop
Analytical literacy for the sector
Digital health is not sustained by buying systems, but by training the people who already understand the clinical record. From OpenSalud LAB to Hazla con Datos, teaching code with real health data.
Hazla con Datos · OpenSalud LAB · LatinR · UNAB
Projects
Selected work
From redesigning a cancer pathway with paper stickers to the natural-language pipelines now running weekly over the country’s surgical waiting lists.
Waiting-list anomaly detection with reverse machine learning
Inverted classification: instead of learning what is typical, the model hunts for what is irregular in the National Waiting List Registry.
- Detección de anomalías
- R
- Python
- Reportería automatizada
Cancer classifier for the national surgical waiting list
NLP, embeddings and gradient-boosted trees to rescue cancer suspicions misfiled as elective surgery. 91.6 % sensitivity over 15,032 clinically validated cases.
- NLP
- Embeddings
- Gradient boosting
- XGBoost
- CIE-10
- Índice de Youden
- Caché por hash
- Human-in-the-loop
Technical positions
Where the usual reading of the problem gets expensive
Positions I hold when designing health systems, and which explain why so many health data projects deliver figures nobody can use.
A waiting list is a censored cohort, not an inventory
Counting who is still waiting describes the survivors of the process, not the demand.
Interoperability is not passing a JSON
Transport is the easy part. The hard part is both sides meaning the same thing.
Chile’s data protection law is an architecture requirement, not paperwork
Treating health data as sensitive data changes the pipeline design, not just the legal annex.
Teaching and community
Methods documented and handed over
2022
Diplomado en Salud Digital, UNAB
Postgraduate teaching in digital health strategy leadership and management.
2024
Hazla con Datos
Data analysis material built on open health data rather than marketing case studies.
2018
OpenSalud LAB
Open health data science bootcamp, with all its material in a public repository.