Context.

I joined Universidad de los Andes in July 2017. The conditions in which I teach, do research and serve today are not those of then.

Four things changed. The students, less than people say. Artificial intelligence, more than anyone foresaw. Research, which moved closer to industry. And the University, which built the structure that made my turn toward teaching possible.

Four fronts, 2017 to 2026.

2017201820192020202120222023202420252026

StudentsI join Uniandes and teach Unit OperationsCurricular reform: Unit Operations becomes a project courseSecond redesign of the courseThe course evaluation recovers

Artificial intelligenceNeural networks to design antimicrobial peptidesChatGPT opens to the public (30 Nov)Training in AI for organizations beginsTask force on teaching in times of AI

ResearchFirst projects as principal investigator, from public callsFirst project with Enel Colombia as principal investigatorA good part of the research is done with companies

The UniversityConecta-TE, since 2012, supports my first redesignDIDACTA brings together the academic support unitsGuidelines for the use of generative AI (October)Comprehensive Development Program 2026-2030 approved (November)

Students

2017I join Uniandes and teach Unit Operations
2021Curricular reform: Unit Operations becomes a project course
2022Second redesign of the course
2024The course evaluation recovers

Artificial intelligence

2020Neural networks to design antimicrobial peptides
2022ChatGPT opens to the public (30 Nov)
2023Training in AI for organizations begins
2026Task force on teaching in times of AI

Research

2019First projects as principal investigator, from public calls
2022First project with Enel Colombia as principal investigator
2024A good part of the research is done with companies

The University

2017Conecta-TE, since 2012, supports my first redesign
2022DIDACTA brings together the academic support units
2024Guidelines for the use of generative AI (October)
2025Comprehensive Development Program 2026-2030 approved (November)

What changed each year in the four fronts this page describes.

The students.

It bothers me when people say that today’s students are not the same as before. What changed was not them: it was their tools and the curriculum with which we receive them. In 2021 the program’s curricular reform turned Unit Operations into a project course, and the evaluations of that course, stable for four years, fell for three years before recovering.

The story of that course, in Teaching

And since late 2022 every student has arrived in class with a language model in their pocket. The surveys I know measure it outside Colombia, but the order of magnitude is clear. Adoption ran faster than pedagogy.

My answer to that is not to say that the students changed. It is to design the course for the student who arrives, with the tools that student brings. That is the idea of the vectors on the home page.

Laptops on the system curve and available NPSH workshop — The system curve and available NPSH workshop, Unit Operations Project, 2026.

86 %of 3,839 students in 16 countries used AI in their studies (2024)Digital Education Council

92 %of students in the United Kingdom used it in 2025; 66 % in 2024HEPI · Kortext

2/3of 95,513 students at 20 public research universities in the United States had used it in 2023-2024Chirikov et al., Science

Artificial intelligence.

ChatGPT opened to the public on 30 November 2022. In 2020 artificial intelligence was, for me, a research method: I used neural networks to design antimicrobial peptides. In less than two years it became a condition of the classroom. A discourse came with it. The most cited reports, almost all from the English-speaking world and from industry, present artificial intelligence as an amplifier of productivity, and the same argument reached higher education.

What the reports promise.

60-70 %
McKinsey Global Institute, 2023
Trillions of dollars a year for the economy, and 60 % to 70 % of employees’ time automatable (a global consultancy, 2023).
75 %
Microsoft · LinkedIn, 2024
75 % of knowledge workers already use AI, and 71 % of leaders would hire someone with less experience and AI skills over someone with more experience and none (31,000 people in 31 countries, 2024).
75 %
Digital Education Council, 2024 · HEPI, 2025
75 % of students believe AI will help them reach their academic goals more efficiently; in the United Kingdom, the most common reason to use it is to save time.

What the evidence shows.

−17 %
Bastani et al., PNAS, 2025
About a thousand secondary-school mathematics students who practised with an unrestricted chatbot improved a lot in practice and scored 17 % below the control group on the exam without AI.
Bastani et al., PNAS, 2025
In the same trial, a tutor that held back the answer avoided much of that harm.
Studies cited in my critical review
The students who depend most on the tool write better texts without learning more, and the same use widens the gap between strong and weak students.
0
My critical review, in final review
controlled studies of learning outcomes with AI in chemical engineering.

The argument is legitimate in a company. In a place of formation it should not be the one in charge, and the evidence explains why. In engineering education there is still no controlled study of learning outcomes with artificial intelligence. That is the gap my critical review, now in final review at Education for Chemical Engineers, points out, and the rule it proposes is the one I defend on the home page: first, do no harm.

Research.

When I arrived, my research was done mostly through projects from public calls: my first projects as principal investigator, in 2019, came from Colciencias and from its successor, the Ministry of Science, Technology and Innovation (Minciencias). Research then moved closer to industry. In 2022 I signed my first project as principal investigator with Enel Colombia, and since 2024 a good part of my research has been done with companies. Since 2023, training in artificial intelligence for organizations has grown every year. That is one of the reasons my research looks more applied today than in 2017.

The projects, in Research

Jesús Rafael Acevedo Mastrogiacomo and Luis H. Reyes feeding the bacterial consortium at the El Paraíso Hydroelectric Power Plant — With Jesús Rafael Acevedo Mastrogiacomo (left), feeding the bacterial consortium we grew for the ecofilter of the El Paraíso Hydroelectric Power Plant, of Enel Colombia.

The University.

Students working at Universidad de los Andes, seen through a window — Students at Universidad de los Andes, Bogotá.

The University also changed, in a direction that made possible what I tell in Teaching. Since 2012 it has had a centre that supports pedagogical innovation with technology, the Centro de Innovación en Tecnología y Educación (Conecta-TE), and later a Centre for Teaching and Learning in the School of Education. My first redesign of Unit Operations, between 2017 and 2018, was done with two Conecta-TE advisers, and the study of the Chemical Engineering Challenge was carried out with me by two researchers from the School of Education.

In 2022 the University brought Conecta-TE and other academic support units together in DIDACTA, which today has an evaluation team that measures before and after each innovation. In October 2024 it published its Guidelines for the use of generative artificial intelligence, to which I contributed from the School of Engineering. And the Comprehensive Development Program 2026-2030, approved in November 2025, placed new educational models and digital transformation among its four directions.

The idea on the home page, that the time artificial intelligence frees up should go back to forming people, goes in that same direction, with one condition: doing it with evidence.