Personal details
Alejandro S.
Summary
I'm (almost, just a few weeks to graduate!) doctor in robotics. For most of my professional life I've researched new methods and algorithms for applying AI planning to robotics. In the world of deep learning, planning is not well-known, but it still has attractive applications, like recycling and healthcare assistance. I have a master in artificial intelligence too, so the new trends of AI are not alien to me.
I've worked in many things: I've implemented machine learning algorithms, and engaged in bot competitions and other forms of competitive programming (CodinGame, HackerRank, Project Euler), which has allowed me to hone my programming skills.
I consider myself to be specially prolific in C++ and Python, which are the languages I use the most in robotics. However, I don't see myself as a collection of languages and frameworks (React + Node.JS + Angular.JS + the whole 100 yards). I see myself as a programmer first, in the sense I like to figure out fast and performant algorithms for certain tasks, and I don't care about the language. I subscribe to Dijkstra's philosophy: "Computer Science is no more about computers than astronomy is about telescopes".
Technical skills
Work Experience
In charge of writing software and solving technical tasks in a robotics laboratory. From time to time, I get involved on a research project. This is a highlight of the most recent project in which I have taken part:
NYAM: robot for assisted feeding: https://www.labora.cat/en-p-ai-eat/, related publication: NYAM: The Role of Configurable Engagement Strategies in Robotic-Assisted Feeding
I've worked as a researcher here as part of my Ph.D. studies. I've worked on coming up with novel strategies for planning and acting in the real world with robots. Although my research was meant to be usable in a wide range of topics, it was channeled mostly through the use case of disassembling electromechanical devices such as hard drives. The framework for this first research was the IMAGINE project.
In the last years of my research I focused more on learning action models from demonstrations. This meant, for instance, learning the rules of a board game just from observing demonstrations of a human playing it. This has great potential because it may enable robots to perform tasks for which it has not been programmed.
Selected publications:
- User Interactions and Negative Examples to Improve the Learning of Semantic Rules in a Cognitive Exercise Scenario [link]
- Online Action Recognition [link]
- Practical Resolution Methods for MDPs in Robotics Exemplified With Disassembly Planning [link]
Full Google Scholar profile: [link]