Castarco

* I have experience with C and C++11, but mainly for toy projects and university homework. I've also have experience with OpenMP (but not with GPU parallellization). I'm used to make and cmake, but autotools isn't strange to me.

The biggest project I worked on is '''Infantium''', a spanish startup aiming to apply machine learning techniques to children's learning process modelling in order to improve and optimize it, helping them to learn more and faster. The idea was to analyse their gaming sessions (with educative games) in order to compute cognitive indicators (numerical coefficients, designed to associate "bigger" to "better") and trying to maximize them. The "problem" isn't one single problem, but many difficult interrelated problems (designing models, guessing models parameters, validating the models, maximizing children's coefficients, dealing with "inflation" and "deflation" problems associated to children scores, designing experiments, making temporal series predictions...).  

This is an interesting project and idea, but we didn't work in a rational way. Sadly, we hadn't psychologists nor neurologists in our team , and our bosses denied the needed resources to experiment and obtain information to validate or refute the models, and to adjust their parameters. They expected magic from their workers: parameterless perfect models (without having experts in the field), or guessing the parameters from divine inspiration. We were three non-expert workers, and many serious companies have more (and more skilled) workers for much more humble projects.

 

I was working in the Data Science team, and I lead it until a more experienced engineer entered into the company. This project made me return to the university. I had to work in very hard problems, but I only solved a few ones. In any case, I've learned a lot of time series analysis, ranking and rating models and artificial deep neural networks.

 

 

My first programmer job was at Bananity (we were programming with PHP, it was another "online social network"), also a spanish startup. Firstly I was frontend developer, and then I jumped to backend development. My first tasks were bugfixing and code style fixing, later I started to work on improving the software's architecture, performance, and security. I also did a DB migration, I had to develop many integrity checking scripts and data transformation scripts. After that, we started working with Vert.x to create a new fast and scalable API.

 

My last 6 months in the company (I stayed there for 14 months) I started working together with the lead data scientist of the company. With him I started working on very simple NLP problems related with DBPedia (we extracted information from DBPedia because our social network was thought to allow people to tell what they love and what they hate, and DBPedia helped us to create those "things" in a simple way).

 

My last project in the company was a short-texts indexer (using MongoDB as storage system, and JBoss as application server). We could have used Sphinx or Solr, but we thought it was more interesting to develop our software (it's open source, is called Platano Indexer). The design (& implementation) of Platano Indexer is far from being perfect, but I've learned a lot doing it and implementing it.

** IRC and mailing lists : I'm very familiarized with mailing lists. Sometimes I use IRC, but less than mailing lists.

** Mercurial or other source code management systems : I'm very familiarized with GIT and many of it's not-so-well-known features. I've also used Mercurial, but I'm not so familiar with it.

** Mediawiki or other wiki software : I'm used to use wikis for professional usage, but also when I edit Wikipedia articles.

** make, gcc, gdb or other development tools : When I program with C or C++ I usually use make or cmake, and GCC as compiler. I'm not very familiar with GDB, I usually find the bugs without having to touch it. I've also use valgrind sometimes to find memory leaks.

 

* What will make you actively stay in our community after this GSoC is over? Well, I think having the sensation that my work is worth, that I'm helping people. Working for startups that only care of trends and don't solve people's real problems isn't for me. I need to solve real problems.

* Did you ever hear about Octave before? Yes.

** When and where? How far have you been involved already? I discovered Octave many years ago, I can't say when exactly, but probably when I started learning ML, or a few months before.

* What was the first question concerning Octave you could not find an answer to rather quickly? When I was trying to parallelize the Zhang-Suen thinning algorithm, and I spent a lot of time to discover how to apply a function to each element of a matrix, but taking into account its neighbourhood. Finally I discovered the colfilt method ^^U.

* I work with Ubuntu, usually the LTS versions. I also have access to Windows 7, Windows 8.1 and Mac OS X Yosemite, but using virtualization.

* I have access to Internet whenever I need it, and I also have "emergency solutions" (3G, friends connections, family connections, or a nearby library with Internet connection).

* I have access to my PC whenever I need it, I have one PC and one laptop, and I also have "emergency solutions".