Collect
Retrieve recent results for configured diseases and Mexican states.
Case study · Jun–Jul 2026
EpiCentro
I built the interactive map end to end—from news ingestion and dataset support to filters, fallback behavior and the state-level interface used to explore epidemiological information.
Overview
Product context and challenge
EpiCentro was created for ITRASIG as a complement to PABLO AI, the initial artificial-intelligence project, continuing the same mission of making epidemiological information more approachable. The wider product combines a news map, vaccination records, dose reminders and short educational activities.
My primary responsibility was the map: collecting and normalizing Google News information, supporting the dataset, exposing it through backend services and turning it into an interactive view across Mexico's states.
Core challenge
News volume is fragmented, repetitive and easy to misinterpret. The map needed to normalize sources, remove duplicates and present relative activity without implying that article counts were official case or incidence data.
The team included 16 people collaborating nationally and remotely. Communication gaps and limited access to the partner during some meetings required clear updates, technical guidance and deliberate integration work across feature teams.
Role
My scope within the team
I worked as a full-stack developer in a 16-person national team. I owned the complete interactive map, contributed to its dataset and backend, supported the unified product integration and advised teammates on technical questions.
- Build the interactive map of Mexico and state-level interactions
- Implement disease and relative-activity filters
- Consume and process Google News information through the API
- Create fallback behavior when live news data is unavailable
- Support construction and normalization of the map dataset
- Develop backend functionality required by the map
- Integrate the map with the complete EpiCentro interface
- Advise team members on technical implementation questions
The complete EpiCentro application was a team outcome. This page focuses on my full-stack ownership of the epidemiological map and my supporting integration and technical-advisory work.
Architecture
How the product layers connected
News sources
Google News and RSS results provide recent articles associated with diseases and locations.
NestJS pipeline
The API normalizes fields, removes duplicates, identifies topics and scores relevance.
Map dataset
Articles are aggregated into consistent state-level records with filterable activity information.
React map
The interface renders Mexico, filters, state details, related articles and fallback data.
Workflow
From fragmented articles to an explorable map
The workflow turns inconsistent external news results into a stable response the interface can explain without presenting it as clinical evidence.
Normalize
Standardize titles, sources, dates and links while removing repeated articles.
Classify
Detect topics, score relevance and aggregate relative article activity by state.
Explore
Filter the map, select a state, review its summary and open the original source.
Contribution
Work completed across the stack
Map interface
- — Implemented Mexico's state geometry and interactive selection
- — Built disease and activity-level filters
- — Created state summaries and related-news navigation
Backend and data
- — Worked on Google News retrieval and API consumption
- — Supported dataset construction and normalization
- — Implemented backend behavior needed by the map
Reliability
- — Added fallback map data for unavailable live sources
- — Kept relative activity separate from official epidemiological claims
- — Connected state records with their original news sources
Team integration
- — Helped unify the map with the complete page
- — Shared technical updates across remote collaborators
- — Advised teammates on implementation questions
Decisions
Presenting news activity without overstating its meaning
Treat news volume as a relative signal
The interface describes article activity rather than cases, incidence, severity or personal health risk.
Normalize before visualization
The backend standardizes and deduplicates external results so the map consumes a predictable dataset.
Design for source failure
Fallback data keeps the map demonstrable when Google News or the processing API cannot return a usable response.
Keep original sources reachable
State summaries lead to related article cards so users can review the publication behind each signal.
Deep dive
The map is a complete full-stack feature, not a static visualization
Data problem
External articles arrive with inconsistent text, dates, sources, locations and repeated results.
Engineering response
A backend pipeline produces state-level records that the React interface can filter and explain.
Product result
Users can explore all 32 states, switch diseases and activity levels, and review source-linked news.
01
State interaction
Hover and selection connect the geographic view with a focused state summary.
02
Filterable context
Disease and activity controls update the information shown without changing the meaning of the signal.
03
Graceful fallback
Demonstration data preserves the primary workflow when live external services are unavailable.
Outcome
What shipped—and what remains unproven
External validation
PABLO AI participation certificate
July 10, 2026
Dr. Octavio P. García González · President, ITRASIG, A.C.
ITRASIG recognized my participation in PABLO AI, its initial AI project. EpiCentro was developed as a complementary product that extends the same public-health information mission.
Read the redacted participation certificateDelivered
- Complete interactive map of Mexico
- Disease and relative-activity filtering
- Google News ingestion and normalized API data
- State-level summaries with related source links
- Fallback behavior for unavailable live data
- Integrated web application prepared for release by the team
Limitations
- No formal end-user, performance or adoption metrics were collected for the delivered version.
- News-derived activity is not official epidemiological incidence and must not be used for clinical or public-health decisions.
- The complete platform was built by a 16-person team; modules outside the map are not presented as my personal implementation.
Lessons
- A public-health interface must communicate the limitations of its data as clearly as it communicates the data itself.
- Large remote teams need explicit technical ownership and consistent progress updates to integrate independent features successfully.
Evidence
Selected product evidence
State-level exploration
Disease and activity filters refine the map while the selected state panel explains the signal and links to its source.
EpiCentro main page
Full resolutionMain page for accessing the map, vaccination records, dose reminders and educational activities.
EpiCentro digital vaccination record
Full resolutionPreview of digital vaccination records area, where users can view and manage their vaccination history.
EpiCentro daily game
Full resolutionPreview of the daily game feature, where users can engage with health-related challenges and activities. It can be a Wordle or a hangman game.
Attribution
Team work, clearly attributed
EpiCentro was completed by a 16-person national team for ITRASIG. The map screenshots and product description are shown with permission. Source code, personal health data, internal datasets and infrastructure remain private.