Dynamic Pricing | Forecasting | Causal Inference | Revenue Optimization | Production ML | Generative AI

Chile | Remote collaboration with US/Canada teams | EST Timezone LinkedIn | Email

Senior Data Scientist and ML Scientist with a 12+ year engineering trajectory across revenue-management, financial services, e-commerce, industrial systems, and applied research.

I specialize in building production systems at the intersection of:

• Dynamic Pricing, Promotions & Revenue Optimization
• Demand Forecasting & Time Series
• Causal Inference & Incrementality Measurement
• Bayesian Modeling & Statistical Machine Learning
• Financial Data Integrity & Decision Support
• Machine Learning Engineering & MLOps
• Generative AI, LLMs, RAG & Agentic Workflows

My focus is not just model accuracy. I build systems where the data is trustworthy, the economic impact is measurable, the implementation is production-ready, and automated decisions remain verifiable by humans.

• US$3M ARR generated through production dynamic pricing models using price-demand elasticity and promotional optimization at LATAM Airlines
• Reduced demand forecasting error from 20% to 9.5% with ensemble time-series models deployed end-to-end on GCP Vertex AI
• Measured promotional incrementality using Difference-in-Differences and quasi-experimental causal inference when traditional A/B testing was not feasible
• Improved Bayesian prospect identification from 43% to 62% with no increase in false positives using a hierarchical PyMC model
• Audited revenue-critical financial and FinOps pipelines through production-data reconciliation, edge-case analysis, regression checks, and freshness controls
• Built predictive optimization, soft-sensor, PINN, anomaly detection, and industrial time-series systems
• Increased e-commerce conversion by 12% through dynamic pricing
• Increased industrial copper yield by 7% through mathematical simulation and process optimization

I design multi-agent workflows for production software engineering and data science, with emphasis on orchestration, task decomposition, controlled autonomy, independent verification, and human-in-the-loop decision making.

Claude Code | OpenAI Codex | OpenCode | MCP | Subagents | Git Worktrees | Tool-Enabled Agents | LLM APIs

Typical orchestration pattern:

Scout → Architect → Engineer → Reviewer → Judge → Human

• Scout: explores repositories, dependencies, data flows, APIs, documentation, and relevant system context
• Architect: designs the solution, defines interfaces, identifies risks, and decomposes work into bounded tasks
• Engineer: performs scoped implementation against explicit requirements and acceptance criteria
• Reviewer: independently analyzes diffs, assumptions, tests, regressions, edge cases, and architectural consistency
• Judge: evaluates implementation evidence and reviewer findings against acceptance criteria, then returns PASS, REVISE, REJECT, or ESCALATE
• Human: retains final authority over ambiguous, high-risk, or production-impacting decisions

I use parallel Git worktrees, subagent delegation, context isolation, model routing, MCP integrations, structured outputs, automated validation, quality gates, guardrails, and escalation rules to improve throughput without treating LLM output as inherently correct.

The objective is not maximum autonomy. It is maximum useful autonomy with inspectability, independent verification, and minimum risk.

My core ML and statistical work includes:

Forecasting | Time Series | Dynamic Pricing | Price Elasticity | Revenue Optimization | Bayesian Modeling | Classification | Propensity Modeling | Churn Modeling | Recommender Systems | Causal Inference | A/B Testing | Difference-in-Differences | Incrementality Measurement | Promotion Optimization | Anomaly Detection | Predictive Maintenance

I also work extensively with mathematical optimization:

MILP | SciPy | CVXPY | PuLP | Inverse Optimization

A recurring theme is connecting predictive models to the actual decision being optimized rather than treating prediction as the final product.

Experience with applied AI systems including:

• Large Language Models and Generative AI
• Retrieval-Augmented Generation (RAG)
• Agentic workflows and multi-agent orchestration
• LLM tool use and Model Context Protocol (MCP)
• Agent-based automation
• NLP and transformer models
• Financial news sentiment and text classification
• Embedding-based retrieval and vector databases
• Human-in-the-loop AI systems
• AI-assisted software engineering

Technologies include:

LangChain | Hugging Face | ChromaDB | Pinecone | Claude Code | Codex | OpenCode

Python | SQL | PySpark | Pandas | NumPy | R

Scikit-learn | XGBoost | LightGBM | PyTorch | TensorFlow | Keras | PyMC

GCP | Vertex AI | BigQuery | AWS | Azure | Docker | Git | CI/CD | MLOps

Claude Code | OpenAI Codex | OpenCode | MCP | LangChain | Hugging Face | ChromaDB | Pinecone

Built production ML systems for dynamic pricing, promotional optimization, elasticity estimation, demand forecasting, customer propensity, churn, ancillary revenue, and causal measurement.

Key results include:

• US$3M ARR from dynamic pricing
• Forecast error reduced from 20% to 9.5%
• Production deployment on Vertex AI
• Difference-in-Differences for campaign incrementality
• Route-level elasticity and MILP-based ancillary revenue optimization

Applied Bayesian modeling, forecasting, financial data validation, MLOps, and agentic software engineering to revenue and financial systems.

Work has included:

• hierarchical Bayesian modeling
• revenue-critical pipeline auditing
• production-data reconciliation
• walk-forward forecasting
• calibrated prediction intervals
• AWS production deployment
• automated business-system integration
• multi-agent engineering in large codebases
• reviewer and judge quality gates

Built systems using:

• Physics-Informed Neural Networks (PINNs)
• inverse optimization
• soft sensors
• predictive optimization
• anomaly detection
• large-scale industrial time series
• predictive maintenance

Developed:

• transformer-based financial news classification and sentiment analysis
• collaborative filtering recommender systems
• personalized Business Intelligence systems

University of Chile (Ranked Top 1% globally and Top 6 regionally by QS Rankings 2026)

MSc, Chemical Engineering, Summa Cum Laude
Quantitative & Industrial Engineering: statistical modeling, mathematical optimization, Industrial process modeling, and data-driven methods
MSc Thesis: Machine Learning Classification for Bioinformatics and Genetic Sequence Analysis, achieving 88.8% sensitivity and 85.5% specificity

Civil Engineer, Biotechnology specialization, Summa Cum Laude

Spanish: Native English: Advanced Professional, IELTS Academic 8.0 / 9.0

Pricing | Forecasting | Revenue Management | Causal ML | Decision Science | Optimization | Production ML | MLOps | Generative AI | LLMs | RAG | Agentic AI | Multi-Agent Systems

Turning complex and imperfect data into production decision systems with measurable business impact.