Implemented an end-to-end forecasting solution for Iconn to predict sales across all 7-Eleven stores in Mexico, significantly improving weekly accuracy compared to the client’s previous model. Additionally, I serve as the owner of the Long-Term Forecasting process, overseeing model deployment, integration into demand planning building-block sessions, and cloud-based automation to ensure scalability and long-term autonomy for the client. PoC of Computer Vision to prove the capacity of the models to capture visual KPIs from a camera that help to monitor the compliance with operational standards on the mine Quebrada Blanca.

Developed and deployed Forecast One on GCP Vertex AI, forecasting sales, transactions, and units across 400+ stores with over 95% accuracy. Later appointed as use-case owner for Merchandise Finance planning. Built a forecasting experimentation platform integrating MLOps and real-world experimentation environments, adopted by Walmart’s Forecast team. Led the development of 180+ agents and 32 agentic workflows validating whether presales proposals met technical and commercial criteria before delivery to clients. The initiative was presented by Globant’s CEO during Globant All Hands 2025, showcasing its strategic value at company-wide level. Designed agents that delivered measurable efficiency gains: STB Agent (Ops): Automated bonus KPI queries, reducing Ops workload by 20%. SOW Validator (Ops): Automated contract review, increasing efficiency by 40%. CV Converter (Capacity): Saved analysts 5+ hours/week via CV normalization. Leadership Agent (People): Multi-agent RAG chatbot answering PTO, benefits, and leadership queries. HR Interviewer Pro (Capacity): Scored candidates from CVs, transcripts, and project data, improving recruitment quality. Finance – Claiming Debt: Automated overdue payment notifications with dynamic tone adjustment and escalation. Delivered a voicebot to handle complaints about unexpected charges using Amazon Connect, Lex, Lambda, Bedrock, and DynamoDB. Delivered initial infrastructure setup for production.

Directed SmartLoader platform ($1M USD budget) for ICV. Managed a team of 5 (Data Scientists + Mechanical Specialists) and delivered a PHM-based machinery health monitoring system. Integrated PostgreSQL DW, RPA/ETL (Airflow), and predictive models with physical priors, generating estimated $277M/year savings and 753 tons CO2 reduction. Designed and implemented an Autoencoder-based health state estimator and conducted causal discovery of subsystem failures for degradation modeling. Reported progress and findings to ICV executives and AC3E board.

Developed high-frequency trading strategies using Bayesian statistics and reinforcement learning. Deployed real-time MLOps pipelines on AWS (EC2, Lambda) integrating Interactive Brokers and CME APIs. Algorithm 'Turing' achieved >100% annual return with 15% max drawdown in backtesting and live trading simulations.

Delivered applied AI consulting for retail forecasting and neural search projects. Supported 3 clients.

Developed high-frequency trading strategies using Bayesian statistics and reinforcement learning. Deployed real-time MLOps pipelines on AWS (EC2, Lambda) integrating Interactive Brokers and CME APIs. Algorithm 'Turing' achieved >100% annual return with 15% max drawdown in backtesting and live trading simulations.

Built anomaly detection routines using time series analysis and neural networks for SAG pumps.