# UPO Digital — Education and Entrepreneurship in the Age of AI

> Education and Entrepreneurship projects with AI at Pablo de Olavide University: learning, venturing and researching with artificial intelligence.

Portfolio of Education and Entrepreneurship projects with Artificial Intelligence by Manuel Chaves Maza, Universidad Pablo de Olavide (Seville, Spain). 27 projects.

- Author: [Manuel Chaves Maza](https://upo.digital/mchaves/) — Universidad Pablo de Olavide
- Web: https://upo.digital/
- Other languages: [Español](https://upo.digital/index.es.md) · [Français](https://upo.digital/index.fr.md) · [Deutsch](https://upo.digital/index.de.md)

## Entrepreneurship — assisted by agents (3)

Three projects that bring AI into business support: a Horizon Europe multi-agent platform, a predictive model of entrepreneurial success and a study of organizations run by agents.

### 001 · MOSAIC

MOSAIC (MULTI-AGENT ORCHESTRATED SUPPORT SPACE FOR EUROPEAN ENTREPRENEURS' INTELLIGENT COLLABORATION) is a two-year Horizon Europe project coordinated by Pablo de Olavide University that proposes an entrepreneurship support platform based on a multi-agent architecture and the Model Context Protocol (MCP). Each agent (market intelligence, business model validator, financing and regulatory advisor) functions as a plug-and-play microservice and is coordinated through an orchestrator that combines their recommendations. The system is complemented by human-in-the-loop governance, where incubator staff supervise and fine-tune the AI to ensure reliable and personalized advice, facilitating the internationalization of startups and reducing information barriers. It will be piloted in various European incubators, with metrics such as entrepreneur satisfaction and investor interest, and will produce a repository of best practices and policy recommendations.

- Open project: https://upo.digital/ssees/index.html

### 007 · AI Bridges Gap

The 'Predictive Edge of AI in Entrepreneurship' page presents a neural network model for predicting the success of new entrepreneurial projects, requiring research access. The model has a layer architecture [50, 25, 12] with rectifier activation and dropout technique, achieving an AUC-ROC of 0.972 and an accuracy of 97.78%. This effort seeks to bridge the gap between AI and entrepreneurship by offering reliable predictions that can guide decisions in incubators and support programs.

- Open project: https://upocuantitativo.github.io/aibridgesGap/

### 020 · Synthetic Organisations

SynthCo Research Manager is the research console of the study on synthetic organisations: agentic companies and organisations (Agentic Organizations and DAOs) that operate mainly through artificial-intelligence agents rather than human staff. The platform structures the work in five phases — sources and setup, data collection, detection and classification, verification, and sample and validation — and adds a dashboard with the data reliability index (source diversity, cross-validation rate, verification depth and sample coverage), the detection funnel from initial scraping to the final sample, and a statistical analysis module against a control group of firms. Identified cases are documented with OSINT and human review, and classifications, reviews and checklists are exported and imported as JSON so the work can move between devices without exposing API keys. It is the infrastructure behind the ongoing paper on the taxonomy, governance and ethics of synthetic organisations.

- Open project: https://upocuantitativo.github.io/ethicagentaisintetic/

## Education — adapted to each learner (11)

Ten platforms and simulators for teaching and studying: personalized itineraries, classroom quizzes, statistics, algebra, mathematical methods and PL/SQL, with AI supporting the teacher, never replacing them.

### 002 · AGORA

The AGORA project focuses on adaptation and optimized generation of learning paths through AI, combining dynamic needs diagnosis, student profiling, and automatic material creation with language models and deep learning. Its main innovation is integrating the recent Nested Learning technique to avoid catastrophic forgetting and reduce the number of required parameters. Objectives include developing learning profiles with a gender perspective, configuring algorithms that reorganize itineraries in real-time, creating a secure local chatbot integrated with a deep learning engine, and designing specialized MCP agents by modality (visual, textual, video) to generate adapted content. Additionally, it proposes measuring the impact on academic performance and motivation through an experimental design.

- Open project: https://upo.digital/agora/

### 004 · InfiniteLearner

InfiniteLearner is an Erasmus+ KA3 educational policy experimentation aimed at creating personalized learning paths for basic competencies (digital, citizenship, literacy, and numeracy) in secondary education and initial VET. The consortium seeks to develop an ethical, inclusive, and evidence-based framework that uses AI without replacing the pedagogical role of teachers, emphasizing the importance of methodological integration and teacher agency. The platform consists of initial diagnosis modules, learning activities, formative assessment, and dashboards that monitor learning signals, as well as AI recommendations where the final decision rests with the teacher. It is structured in work packages for management, methodological co-creation, international pilots, consolidation, and transfer of results.

- Open project: https://infinitelearner.eu/

### 005 · Quiz Platform

This online quiz application allows students to select their plan, class (EPD), and role to participate in real-time assessments. It offers traditional test modes, MCP tests, and specialized tests, recording correct/incorrect answers and displaying complete scores and final rankings. It also incorporates surveys for teachers to collect feedback, with participation statistics visible in real-time. The platform provides information about personal data protection and allows users to exercise their rights of access, rectification, and deletion by contacting the academic responsible.

- Open project: https://upo.digital/app/

### 006 · YourLearningWay

YourLearningWay is an intelligent system that designs personalized learning itineraries according to student profile. It offers different profiles: students (with intelligent weekly planning, task management, reminders, grade tracking, and AI chat), tutors/parents (real-time monitoring and performance alerts), and teachers (interactive quiz creation, CSV import, rankings, and AI-assisted evaluation). It has Free, Lite, and Premium plans with trial periods and scalable services such as unlimited chat, advanced statistics, and API integration. The project is based on the 'Neural Network Learning' research from UPO, selected as a pioneer project, and aims to be a tutor adapted to each learning moment, detecting difficulties, demotivation, and allowing teachers to know in real-time the degree of comprehension. It is implemented on the FIWARE platform, with learning profile databases and the ability to integrate with LMS such as Moodle or Chamilo.

- Open project: https://yourlearningway.com/

### 012 · Hypothesis Testing

This interactive tool illustrates hypothesis testing for a population mean, showing how type I error (α), type II error (β), power, and confidence vary. It starts from a null hypothesis μ₀=30 versus an alternative μ₁=40 with a known standard deviation and n=25, and calculates the distribution of the sample mean, showing the percentages of errors and correct decisions in a contingency table. Interpretations are explained: type I error is rejecting H₀ when true, type II error is not rejecting it when false, and it highlights how increasing α, sample size, or the difference between means reduces β. This simulation helps understand the relationship between parameters in hypothesis test design.

- Open project: https://upocuantitativo.github.io/hipotesis/

### 013 · Statistics

This website gathers practical cases, activities, and interactive examples to facilitate the understanding of descriptive statistics and its application. Created by Manuel Chaves from the Quantitative Methods area of the Economics Department at Pablo de Olavide University, it is designed to support learning in subjects such as GADE and double degrees through content organized by topics and a 'learning by doing' approach.

- Open project: https://upo.digital/estadistica/

### 016 · MotivationLearn

MotivationLearn / OULAD Explorer is a learning analytics project aimed at investigating student motivation and its improvement through reinforcement learning. It uses the OULAD dataset, with 32,593 students, 22 module editions, and 10.6 million interactions in the virtual environment, transforming them into student profiles, predictive models, and 'virtual clones' on which agents are trained to learn which personalized educational interventions work best.

- Open project: https://upocuantitativo.github.io/motivationlearn/

### 021 · AlgebraAI

algebraai is an interactive applied-algebra website, with AI-assisted generation and solving of exercises, built as teaching material for Mathematical Methods. It covers two blocks. The first, linear codes over F2: definition of an [n,k,d] code, generator matrix G and parity-check matrix H in systematic form, minimum distance and linearly dependent columns, correction and detection capability, syndrome decoding through the coset leader, and the Hamming bound with the table of perfect codes; it includes an H and G builder, an encoder with a noisy channel and a step-by-step decoder. The second, QR factorisation: fundamentals and uniqueness of A=QR, the three classical algorithms (classical Gram-Schmidt, modified Gram-Schmidt and Householder reflections) solved step by step, application to least squares, and a numerical comparison of stability and conditioning. It is completed with a test suite for self-assessment.

- Open project: https://upocuantitativo.github.io/algebraai/

### 022 · Mathematical Methods (MMI)

Mathematical Methods for Engineering gathers the teaching materials of the course in the Computer Engineering degree at Pablo de Olavide University, organised in four blocks. Block 1, error theory: absolute and relative error, precision and accuracy, significant digits, error propagation, IEEE 754 arithmetic, conditioning and numerical stability. Block 2, non-linear equations: root location with Bolzano, bisection, secant, regula falsi, fixed point and Newton-Raphson, with a convergence comparison. Block 3, linear systems and interpolation: solving A.x=b by Gauss, LU, Cholesky, Jacobi, Gauss-Seidel and relaxation, and polynomial interpolation with Lagrange, Newton divided differences, Hermite and cubic splines. Block 4, numerical differentiation and integration and ordinary differential equations: finite differences, Newton-Cotes formulas, composite trapezoidal and Simpson rules, one-step methods, Adams-Bashforth, Adams-Moulton and predictor-corrector schemes. Each block combines theory, worked examples and computation notebooks.

- Open project: https://upocuantitativo.github.io/MMI/

### 023 · Interactive PL/SQL

Interactive PL/SQL is the study environment for Oracle's procedural language in the Database Architecture course. The syllabus is organised in seven practicals plus a concurrency topic: introduction to PL/SQL (blocks, data types, variables, %TYPE, %ROWTYPE and SELECT INTO), control structures, SQL functions and cursors, error handling, procedures and functions, packages, general review and concurrency. Every block can be worked through in seven different modes — theory, quiz, fill in the blanks, what is missing?, reorder the code, what would you change? and exam — designed to move the student from recognition to writing real code. The application keeps a cumulative global score and per-practical progress, so students can spot their gaps while practising with subprograms, packages, cursors and triggers.

- Open project: https://upo.digital/plssql/

### 027 · Interactive Java Collections

Interactive Java Collections is the practice trainer for the Data Structures course, built for an exam in which students receive a problem statement and must implement it in Java. It gathers the experiments, exercises and problems of practicals EPD-1 (Java Collection Framework: Collection, generics, toArray, Iterator and safe removal) and EPD-2 (List, Collections, Comparable and Comparator, ListIterator, Set, HashSet, TreeSet and SortedSet), plus a block on hand-made linked lists. Every task is solved in a code editor and a static checker grades it instantly: a mark out of 10, why not (missing requirements and how to fix them), typical serious errors (ConcurrentModificationException, comparing strings with ==, equals without hashCode, double next or remove), why yes and recommendations. It includes three-level hints with a penalty, compiled model solutions with their real output and explanations, predict-the-output questions, a recommender that detects the weakest concepts, an exam cheat sheet and a timed mock exam that prioritises the student's mistakes.

- Open project: https://upo.digital/estructuras/listas/

## Research — open and reproducible (12)

Twelve applied studies built on open data and interpretable models: gender and development, algorithmic ethics, criminology, tax policy, rural tourism, Alzheimer care, human trafficking and territorial planning.

### 008 · AI for Development

This interactive dashboard analyzes GDP per capita growth in 52 low or middle-income countries using 131 gender and development indicators. Models tested include Random Forest, XGBoost, neural networks, and Gradient Boosting; Random Forest was the best with R² = 0.760 and RMSE = 12.60. The methodology includes correlations for variable selection, recursive hyperparameter optimization, and validation through cross-validation and bootstrap. The analysis provides model rankings, identification of key variables through SHAP values, and robustness diagnosis with learning curves and residual analysis. Results facilitate policy recommendations on the relationship between gender and development indicators and economic growth.

- Open project: https://upocuantitativo.github.io/aidevelopment/resultados/dashboard_complete.html

### 009 · Ethicalgorithm

Ethicalgorithm is an open computational framework for evaluating the moral consistency of language models through adaptive multi-turn dialogues. Its key innovation consists of applying adaptive counterarguments: if a model adopts an ethical stance, pragmatic pressures are introduced, and if it is permissive, moral objections are raised, thus revealing opportunistic reasoning. Validation with an independent panel of experts shows 93.8% agreement between algorithm and evaluators and a Cohen's κ of 0.728. The protocol is based on three axioms (measurement through dialogue, pressure that reveals value hierarchies, and adaptive detection), implements a four-step loop (dilemma presentation, adaptive probing, pressure escalation, and critical choice), and generates scores in ethical awareness, consistency, ethical priority, and contradiction avoidance.

- Open project: https://upocuantitativo.github.io/ethicalgorithm/

### 010 · AgriculturaI

Artificial intelligence platform for farmers. Disease diagnosis, crop management, and technical assistance at your fingertips. What can you do with AgriculturaI? Visual Diagnosis: Photograph your plant and discover what disease it has in seconds. Crop Management: Register and receive personalized recommendations for your crops. Virtual Assistant: Consult about phytopathology, nutrition, and irrigation in real-time. Agricultural Products: Find and compare fertilizers, fungicides, and more.

- Open project: https://upocuantitativo.github.io/agriculturai/

### 011 · AI Gender for Development

This 'Dual Target Analysis' study simultaneously predicts GDP per capita growth and ease of doing business score from 128 gender and development indicators in 53 countries. After removing mortality variables, the 15 variables with the highest correlation are selected and models (Random Forest, XGBoost, Gradient Boosting, and neural networks) optimized by grid search and validated with bootstrap are trained. The fiscal score model achieves R²≈0.97 and maintains positive performance in cross-validation, demonstrating that tax variables, contraceptive prevalence, judicial fees, and fertility are key predictors. The analysis includes correlations, clustering by countries, and robustness evaluation, and highlights policy implications on how gender and development indicators influence the business environment.

- Open project: https://upocuantitativo.github.io/aigenderfordevelop/

### 014 · Criminology Simulator

This website is an interactive simulator created by Manuel Chaves (Quantitative Methods area, Pablo de Olavide University) that allows modifying socioeconomic variables in real time, exploring scenarios (optimistic/realistic/pessimistic), visualizing correlations, temporal clusters, and comparing predictive models (e.g., regression, Random Forest, Gradient Boosting), with metrics and interpretation to support data-driven decision making. It also includes a downloadable dataset and reports, making it directly reusable as a basis for undergraduate theses, master's projects, analysis/model validation practices, and policy/scenario design exercises.

- Open project: https://upo.digital/simulador_cri/

### 015 · S-Cultural

This project focuses on using music as a tool for social inclusion and community development in Polígono Sur (Seville). Specifically: fostering local musical talent by identifying and training children, young people, and adults with musical interest in Polígono Sur; promoting social cohesion and self-esteem through collective creative experiences (percussion bands, rap, singing, etc.) that reinforce personal confidence and neighborhood pride; breaking stigmas and empowering the community by improving the neighborhood's image through visible cultural products (concerts, recordings) with local protagonists; and strengthening the existing cultural network by building alliances between UPO, neighborhood associations, the Factoría Cultural, and other agents to ensure project continuity. These objectives are inspired by the philosophy of inclusive culture: integrating art into the daily reality of Polígono Sur to transform unfavorable social dynamics.

- Open project: https://upo.digital/s-cultural/

### 017 · ARIADNA

ARIADNA (Assistance, Referral, Inclusion and Autonomy through Digital Navigation and Access) is a transnational project aimed at developing a digital navigation and support system for victims and survivors of sexual exploitation and human trafficking in Europe. It functions as a 'digital thread' that helps confidentially, multilingually, and accessibly navigate the complex network of legal, social, health, and employment services, also integrating professional training, cross-border cooperation, prevention actions, and evidence generation to strengthen European victim protection policies.

- Open project: https://upo.digital/ariadna/

### 018 · APOYO

APOYO is a confidential digital assistance tool designed to provide immediate and accessible guidance to people in vulnerable situations, especially linked to contexts of trafficking, exploitation, or need for specialized support. Its approach combines informational assistance, resource navigation, secure communication, and referral to professional services, with an interface designed to overcome language barriers, facilitate access to help, and act as a clear, discreet, and useful entry point.

- Open project: https://upocuantitativo.github.io/apoyo/

### 019 · TipoImpositivo

TipoImpositivo is an interactive fiscal policy simulator that allows real-time modification of personal income tax brackets and corporate tax rates to analyze how different scenarios affect public revenue, citizens according to their economic profile, and costs borne by businesses according to their size. The tool incorporates predefined scenarios, individual calculations for specific incomes, and several comparative tabs that facilitate studying the distributive effect, progressivity, and effective impact of different tax reforms.

- Open project: https://tipoimpositivo.onrender.com/

### 024 · Tourism: Efficiency & Accessibility

Tourism research line combining efficiency and accessibility. Its centrepiece, European Rural Tourism Efficiency, is a reproducible framework that estimates the efficiency of European rural tourism through interpretable machine learning and decomposes its country-level determinants. The methodology chains ten phases: data ingestion (Eurostat panel, World Bank API, UNESCO and Factbook), structural extension with 23 country-level indicators, DEA frontier (CCR, BCC, super-efficiency and SBM), censored Tobit regression with MLE, GMM and Simar-Wilson bootstrap, six machine-learning estimators plus ensemble (Elastic Net, Random Forest, GBM, XGBoost, LightGBM and stacking), explainability through permutation, SHAP and PDP, a comparative scoreboard across DEA, Tobit and ML, scenario projection on choropleth maps, an out-of-sample predictor for 31 countries outside the study, and a final bilingual report. The line is completed by an AI tourism agent and the ATP-PAI project on accessible tourism powered by personalised AI.

- Open project: https://upocuantitativo.github.io/deatobit/

### 025 · Alzheimer Chat-RL

Alzheimer Chat-RL is a research prototype from Pablo de Olavide University, in the Knowledge Engineering area, aimed at intelligent anticipation in Alzheimer's care through conversational artificial intelligence and reinforcement learning. A single platform offers a different interface for each audience. The patient talks to Aurora on a warm, jargon-free screen, with a large microphone, spoken replies, music, photographs and sayings, and a one-touch call to a relative. The clinician has a panel with patient triage by urgency, longitudinal progression, MMSE and anxiety tracking over time, reviewable transcripts, clinical notes per session and configurable alert thresholds. The research team accesses dense telemetry with the reinforcement-learning policy, rewards, cognitive domains, training curves and an experiment comparison against the baseline. The relative receives simple supervision, with a traffic-light status, alerts translated into plain language and a summary of every conversation.

- Open project: https://upocuantitativo.github.io/alzheimer-chat-rl/

### 026 · SDAU Doukkala 2045

Integrated territorial platform for the Urban Planning Master Scheme (Schema Directeur d'Amenagement Urbain) of the Doukkala coastline (Kingdom of Morocco, Casablanca-Settat region), with a 2045 horizon. It brings together in a single dashboard the master scheme's information for 16 municipalities — 3 urban, 12 rural and 1 mixed — 207,687 hectares of study perimeter and 130 kilometres of coast, with a population of 624,069 inhabitants according to the 2014 RGPH census and a projected 40% growth by 2045. The platform includes a territorial GIS, the portfolio of 14 structuring projects (with the Jorf Lasfar industrial hub and the Mazagan Beach Resort among the major operations), a battery of territorial indicators, predictive models and prospective scenario simulation, plus the governance modules: financing, partners, transparency and the 2024-2045 timeline. It is available in French, English, Spanish and Arabic.

- Open project: https://upo.digital/sdau/

## Author profile

### 003 · Manuel Chaves Maza

Manuel Chaves Maza, creator and promoter of Upo.digital, professor and researcher in Applied Quantitative Methods at Pablo de Olavide University. He presents his academic profile, links to ORCID and research networks, and highlights publications on defining entrepreneurial success, evaluating entrepreneur performance, and territorial intelligence systems. The news section includes awards such as the First Thesis Prize 2020 from the Centro de Estudios Andaluces and recognition from the Junta de Andalucía for his research in AI applied to entrepreneurship. The projects section mentions an AI system to advise entrepreneurs through success prediction and personalized recommendations, and an intelligent educational platform that uses neural networks to optimize learning itineraries and integrates with LMS. It also includes contact information and quick links to his academic profiles.

- Open project: https://upo.digital/mchaves/

---

For AI agents: https://upo.digital/llms.txt · https://upo.digital/sitemap.xml · https://upo.digital/.well-known/ard.json