Francisco Pérez-GalarceMachine Learning · Research

Research into practice

From analytical models to working decision tools

Advanced analytics creates value when rigorous methods become usable products. These examples show how predictive and optimization research can be transferred into tools that support operational and policy decisions.

Product examples

Two complementary technology families

01

Predictive analytics

Predictive Intelligence Platform

A configurable product for estimating risk, forecasting demand, and identifying early-warning signals from operational and historical data.

Transfer output Model service + monitoring dashboard

Core capabilities

  • Classification, forecasting, and time-to-event prediction
  • Uncertainty, calibration, and data-shift assessment
  • Fairness diagnostics and explainable predictions
  • Model monitoring and periodic retraining workflows

Application examples

  • Patient and graft survival after transplantation
  • Emergency-unit severity and length-of-stay forecasting
  • Customer churn and energy-demand prediction
  • Agricultural disease and production early-warning systems
02

Optimization models

Optimization Decision Engine

A scenario-based decision-support product that recommends allocations, schedules, routes, or team configurations under operational constraints and uncertainty.

Transfer output Optimization API + scenario workspace

Core capabilities

  • Resource allocation and scheduling
  • Network design, routing, and facility location
  • Multi-objective and fairness-aware optimization
  • Robust planning under uncertain conditions

Application examples

  • Kidney exchange and health-resource allocation
  • Collaborative learning team formation in schools
  • Vaccination-center and emergency-shelter planning
  • Agricultural harvest and warehouse operations

Transfer pathway

A practical route from problem to pilot

  1. 01

    Frame the decision

    Define the operational problem, users, constraints, and success criteria.

  2. 02

    Assess data readiness

    Evaluate available data, quality, governance, bias, and integration needs.

  3. 03

    Build and validate

    Develop a reproducible prototype and compare it with current practice.

  4. 04

    Pilot in context

    Integrate the tool into a real workflow, monitor outcomes, and prepare scaling.

Possible deliverables

Designed for adoption

Proof of concept Decision-support web application Model or optimization API Reproducible analytical package Documentation and training Impact-evaluation framework

Collaboration

Have a decision problem worth translating?

I am interested in collaborations where advanced analytics can move from a research result to a tested, useful technology.

Discuss a technology-transfer opportunity