15 years across mining, forestry, retail, tech, and higher education — from data collection & standards mapping for MBA accreditation (AMBA & AACSB) to ML systems that drive revenue, reduce risk, and retain customers. Industrial Engineer since 2013 — thesis: GJR-GARCH volatility forecasting on live market data.
Certified in Data Science — Alkemy · Talento Digital/SENCE (May 2026)
My background is the differentiator: I connect models to business decisions, not just technical outputs.
I work on problems that move the business:
| Domain | What I solve |
|---|---|
| 📈 Revenue Growth | Which customers to target, when, and with what offer |
| 🔄 Retention | Why customers churn and what intervention changes that |
| 💰 Cost Reduction | Where operations waste resources that data can recover |
| 🔮 Forecasting | Demand, revenue, and risk — before the quarter closes |
| ⚙️ Operations Efficiency | Bottlenecks, throughput, and process optimization |
| 🛡️ Risk Control | Fraud detection, anomaly signals, threshold calibration |
| 🏷️ Pricing | Volume vs margin tradeoffs backed by behavioral data |
| 🚚 Supply Chain | Allocation, inventory, and distribution decisions |
Projects that start at "model" look junior. Projects that start at "business problem" look senior.
📍 Villarrica, Chile · Open to remote roles globally 🌎
Core
Data & Viz
Apps & Deployment
Methodology
Every project follows CRISP-DM as the structural backbone — because it forces the right question first: what business decision does this support?
Lean eliminates waste at every phase:
- No exhaustive EDA — only what informs the decision
- No over-engineered models — simplest model that solves the problem
- No technical-only output — every project ends with a business-readable conclusion
Agile keeps delivery honest — short sprints, working artifacts, no analysis paralysis.
The analytical chain always runs:
Business Problem → Decision to Support → Data → Model → Business Impact
Projects that start at "model" look junior. Projects that start at "business problem" look senior.
| Repository | Focus | Highlights |
|---|---|---|
| applied-data-science-portfolio | 🚀 Applied Data Science Portfolio | Business-focused ML · Fraud Detection · Continuous growth |
| bootcamp-data-science-portfolio | 🎓 Foundation Portfolio | Delivered projects · Guided case studies · Analytics fundamentals |
Numbers documented in the repo — defensible in any interview.
| Business Problem | What I Did | Result |
|---|---|---|
| Fraud systems miss fraud AND block good customers | Cost-optimized Random Forest calibrated at t*=0.0414 | 81.7% fraud caught · 79.7% Expected Loss reduction · USD 16,980 saved |
| Retention risk invisible to the team | Statistically confirmed churn 38% above industry benchmark | 41.4% churn · z=5.18 · p<0.001 |
| Marketing spend with no targeting model | Built spend prediction to enable 3-tier customer segmentation | R²=0.977 · MAPE=3.7% (GradientBoosting) |
| No behavioral visibility across customer base | Segmented full customer base into actionable clusters | 8,068 records · KMeans + PCA + t-SNE |
| Inventory & branch allocation without data | Distributed queries on full transaction history at scale | 1M records · Spark SQL · Catalyst optimizer |
| No analytical foundation | End-to-end retail data pipeline from raw to insight | $37.8M CLP in transactions processed |
- ☁️ Cloud pipeline — production-scale Spark on AWS EMR Serverless + S3
- 🤖 Credit Risk Agent — CMF public data · AI agent for credit market analysis · early stage
