Machine Learning applied to Robotics

Neural network language model

Second year computer science master research specialized in data analysis and machine learning. My internship is supervised by Benjamin Piwowarski and Sylvain Lamprier. We experiment how parametrized neural network language models can improve information retrieval systems.

In charge of developing and running the experiments in the field of Machine Learning using HTK to implement the Hidden Markov Model. I was supervised by Pr. Thierry Artieres.

In charge of the Rails servers development and customer demonstrations.- Test the product to ensure that it operates satisfactorily.- Specify and ensure backward-compatibility of the REST API.- Handle support and feedback.- Designe databases.- Lead and participate in application testing.- Participate in code reviews, create and practice coding standards.

In charge of the Linux based OS embedded (framework, end-users applications, tests, facilities) in all devices, and the desktop software (simulator, update tool, IDE). Improve the internal development processes and environment. Lead the R&D department.- Improve the simulator: introduction of QEmu ; new communication architecture between the IDE and the simulator ; abstraction layer over the kernel for emulation.- Supervise a contractor for the development of the update software : validate the architecture and the development for Windows ; coordinate the contractor with the internal teams ; ship releases to customers.- Set up an automated test bench for hardware and low-level software : designed the test specification language in Ruby ; architecture the test environment based on RSpec.- Reorganize process development : increase bugs tracker ; improve version tracking ; upgrade development branch management policy ; developed tools for code review (Shell Git hooks).- Developed new embedded graphics engine in C++ : optimize rendering pipeline ; write code generator integrated in application development IDE ; write automated test framework.- Write development guidelines.- Managed a team of 2 R&D engineer and several interns.

In charge of the development of a tool suite called Gostai Suite designed to help developing robots behaviours in Urbi.- Developed Gostai Studio and Gostai Lab in C++/Qt (http://www.gostai.com/products/studio/). I was the only developer of these big projects helped sometimes by interns. I designed the user interface with the graphic designer. The software architecture was oriented to improve code reuse and testing.- Designed a software client for an experimental pilot with French telephony operators. The goal was to be able to remote control robots. I wrote the protocol to enumerate and control robots. I also wrote a tool to configure the robots' network environment.- Did the software demonstration for the investors.- Tested and deployed the tools on Window, Mac OS X and Linux.

Involved in the Kodamat project as an intern developer. The goal was to integrate bleeding edge IA system based on reinforcement learning in the Counter Strike project to control a group of bots. I was supervised by Oliver Sigaud and Thomas Degris.Also:- lead a team of 4 others interns.- helped in developing the components system in C++/Java of the laboratory.

- Developed of a tool to test and grade more than 300 students' program every weeks.- Wrote a performance regression tracker system.

Helped to administrate the computing cluster and I wrote a molecule rendering tool in C using GTK, OpenGL and POV-Ray.