Justin Arce

AI/ML Engineer with a Master's in Big Data and experience delivering production ML, GenAI, and end-to-end cloud analytics systems. Built revenue-focused churn modeling and LLM-enabled decision support, production feature infrastructure, and Azure Databricks analytics spanning batch and streaming ingestion through business dashboards.

Combines

Python, SQL, experimentation, and data engineering with measurable commercial outcomes, including $80,000 in annual savings and a 300% increase in revenue recall.

AI/ML Engineer at Rocket Software (2026-05 – Present)

• Production Decision Support: Implemented LLM-enabled explainability for production churn models, translating SHAP-based model output into plain-language recommendations for sales representatives and supporting revenue-retention decisions.
• LLM Quality Improvement: Applied DSPy to systematically optimize the prompt pipeline for model explanations, improving consistency and reliability of generated recommendations.
• ML Platform Migration & Cost Reduction: Led the migration of the ML modeling practice from Dataiku to Snowflake, rebuilding the feature and model pipeline and generating $80,000 in annual savings.
• Commercial Model Scaling: Expanded the churn-prediction platform from two to multiple models across products, increasing revenue recall by 300%.
• Feature Infrastructure: Built a production Snowflake feature store with 350+ features, monthly SQL-procedure refreshes, and a view-based inference layer supporting the full modeling practice.

Senior Process Analyst (Mapper) at Mimica (2024-05 – 2026-05)

• LLM-Structured Data Pipelines: Architected an end-to-end ETL pipeline transforming raw Graph JSON into LLM-ready structured formats, automating complex process-insight documentation at scale.
• Automation & Business Impact: Designed LLM prompts combined with statistical clustering for automated process discovery, accelerating delivery cycles by 20-30% and saving 2-4 days per two-week project.
• Evaluation & Validation: Served as Human-in-the-Loop technical lead, validating mapping accuracy against high-volume activity logs and producing ground-truth data used to fine-tune proprietary ML models.
• Algorithm Engineering: Reverse-engineered process-mining logic and developed a custom greedy algorithm with dynamic volume-coverage parameters that outperformed the legacy solution.

Trader and Financial Markets Analyst at MBCC Advisors (2022-02 – 2023-05)

Process Automation Analyst at Universidad de Costa Rica, Public Administration School (2021-08 – 2021-12)

Quality Management Intern at PRUMISA (2021-03 – 2021-07)

Master's in Big Data & Data Engineering – Universidad Complutense de Madrid (UCM)

Bachelor of Science in Industrial Engineering – University of Costa Rica

Certificate (In Progress) in Agentic AI – Johns Hopkins University

Specialization in Data Science – Tecnologico de Costa Rica (TEC)

Specialization in Design of Experiments – Arizona State University