Maximiliano
Gómez Mallo

Taqueando

Order and cash-management system for a family restaurant in La Paz, in production since mid-2025. It covers the whole shift: kitchen tickets, payments, cash-drawer reconciliation and per-user audit logs. I deploy and maintain it myself on a VPS.

• Node.js
• Express
• React
• MySQL

• ~7,950 orders
• 246 operating days
• in production since 2025

El Observador Digital

Daily pipeline that scrapes 6 Bolivian news outlets, groups articles into events by semantic similarity and measures editorial bias by comparing how each outlet covers the same story.

• Python
• FastAPI
• PostgreSQL
• MiniLM
• Gemini

• 6 outlets
• cosine > 0.65 + Union-Find
• 2 daily jobs

bioAlerta

Model that estimates whether a Bolivian bird or mammal species is threatened, from biological traits and geospatial features, to prioritise the 175 species the IUCN has not yet classified.

• Python
• scikit-learn
• GeoPandas
• DVC
• MLflow

• ROC-AUC 0.848
• 1:34 class imbalance
• 2.2 GB versioned

Taqueando Analytics

Forecasting the restaurant's daily revenue on its real sales. I compared SARIMA, Holt-Winters and an LSTM; the seasonal period is 5, not 7, because the place opens Tuesday through Saturday.

• Python
• statsmodels
• TensorFlow
• Flask

• MAPE 22.0% (LSTM)
• vs 27.3% and 35.2%
• 246 operating days

Tráfico LP

Unsupervised analysis of road congestion in La Paz and El Alto using a dataset I collected myself, hour by hour, through the Distance Matrix API over the OpenStreetMap road network.

• Python
• scikit-learn
• UMAP
• OSMnx
• Folium

• 96 readings × 250 segments
• 10.8% of arteries
• concentrate the congestion

wavegame

Real-time multiplayer party game: one player sees where the target falls on a dial and gives a single clue for their team to guess it. Room state in Redis and a service layer kept separate from the network.

• Node.js
• Socket.io
• Redis
• MariaDB
• Docker

• real-time rooms
• ~50 socket events
• 3 game modes