University of Cambridge, Department of Engineering, Oct 2005 - Now
Ph.D. in Computer Science, Machine Learning.
My advisor is Zoubin Ghahramani

Master of Engineering in Computer Science, Cornell University, June 2003
GPA = 3.92 / 4.0

B.Sc. in Computer Science with Minor in Applied Math, Cornell University, June 2002
Dean's list Spring 1999 and Fall 2002

Math Sup/Spé, Lycée Polyvalent Ort, Strasbourg, France, Sept 1997 - Jan 1999

Scientific Baccalaureate with a Mention, Lycée International, Strasbourg, France, July 1997

Graduated with highest Honors, Mayfair High School, Los Angeles, June 1995
Exchange student in an American high school, scholar athlete award.

June 2009 - Sept 2009
Microsoft Research, Redmond
Intern at Microsoft Bing working on ranking of web search results. Ensemble methods were performed on large library of models and on datasets of millions of documents. Our method showed a significant improvement of NDCG and is currently being tested for production use.

May 2006 - Sept 2006
University of California, Berkeley
Visiting Researcher, worked on text mining and search interfaces.
BioText Search Engine: beyond abstract search published in Bioinformatics.

June 2005 - Oct 2005
University of California, Irvine
Institute for Genomics and Bioinformatics
Visiting Researcher, worked on kernel methods for the analysis of chemical information.
One- to Four- Dimensional Kernels for Small Molecules and Predictive Regression of Physical, Chemical, and Biological Properties published in JCIM.

June 2004 - Sept 2004
Intern, Xerox Research Center Europe, Grenoble
Here is a presentation of Ensemble Selection given at Xerox.

• My master's final project: Ensemble Selection from Libraries of Models is published in ICML 2004.
• Teacher Assistant on CS 478: Machine Learning at Cornell.
• BOOM 2003 on Extreme Ensemble Selection

Sept 2002 - Dec 2002
Teacher Assistant on CS 100: Intro to Computer Programming at Cornell

June 1999 - June 2002
Math tutors for students taking the scientific baccalaureate.

Mathematics: Multivariable Calculus, Analysis, Linear Algebra, Abstract Algebra, Algorithmic.

Programming: Prefered programming language is Python. I also like to program in C, Java and Scheme. Fluent in Unix tools. Other skills in HTML, CSS, PHP and SQL.

Language: French (native) and English (fluent)

Hobbies: Tennis, Squash, Golf, Sailing, French Literature, Art, History, Travelling

Cloud Mining

Cloud Mining automagically builds cool faceted search interfaces for your data. Simply get your data in a specific format, provide a custom look and feel for each search result and Cloud Mining does the rest.

See it in action on IMDb or on DBLP.

Biomed Search

Biomed Search is the largest search engine to look up images specifically in the biomedical domain. The system searches within captions and referring texts to images. Currently over one millions images are indexed. The projects was created in 3 months of works. I used the Lucene retrieval engine and the web.py framework. I did everything on the site: setting up the server, parsing, indexing and web programming. All the back end work was programmed in python.

Here is a presentation of Biomed Search given at the University of Cambridge.

Biotext Search Engine

I also contributed to UC Berkeley's Biotext Search Engine. Search within captions and abstracts of biomedical images. More information on this publication.

MLSS Admin - [source code]

The admin we used at Cambridge for the summer school. It let us browse through the fast amount of candidates, comment on each applicant and assign scores. Try a demo, how to use it, get the source code on github. Feel free to use for subsequent summer schools.

Also available is a simple mass mailer (try a demo, source code) used to email all the admitted students.

The admin has been succesfully used to organize the Machine Learning Summer School 2009 and 2010 so far.

Chiefmall

With Chiefmall you can search for a contractor near you. Over 420 000 contractors including all active licensed contractors in California are currently listed. Contractors who join in can easily build up their own website with a unique url, picture portfolio, multiple work locations and more. Users can contact these contractors or post jobs to them.

Here is an example with explanation of a custom website. And here are some screenshots of a contractor's account: dash board with todo list, describe your company, service locations, the portfolio, adding a caption, get jobs or get your license verified.

It took me 6 months of work to create this project. I did everything on the site: setting up the server, crawling parsing the data, geocoding, UI, MVC design, web design, concepts and ideas and SEO. All the back end was programmed in python.

The website also features an admin created using a program that makes on the fly Gmail like looking interface. Some screenshots: the state categories, and what they are referring to, approving or banning companies, search logs or support with knowledge database.

Google Modules- [source code]

A iGoogle gadget directory released months before Google released their own. The site receives about 25 gadget submissions per day and has over 1700 del.icio.us bookmarks. Google Modules was co-produced with my friend Philipp Lenssen.

A python rewrite of Google Modules is available open source for education purposes.

Wikitrivia - [source code]

Random automatic quizes using wikipedia. An interesting experiment relased before the word mashup was first coined and which received over 50,000 hits on the first three days it launched. WikiTrivia is now open soure.

COURSES

Relevant machine learning courses I took at Cornell:

Marti A. Hearst, Anna Divoli, Harendra Guturu, Alex Ksikes, Preslav Nakov, Michael A. Wooldridge, and Jerry Ye. BioText Search Engine: beyond abstract search, Bioinformatics, 2007
[pdf]

Chloé-Agathe Azencott, Alex Ksikes, S. Joshua Swamidass, Jonathan Chen, Liva Ralaivola, and Pierre Baldi. One- to Four- Dimensional Kernels for Small Molecules and Predictive Regression of Physical, Chemical, and Biological Properties, Journal of Chemical Information and Modeling (JCIM), 2006.
[html]

Rich Caruana, Alex Niculescu, Geoff Crew and Alex Ksikes, Ensemble Selection from Libraries of Models, Proceedings of the 21st International Conference on Machine Learning (ICML), 2004.
[ps] [pdf]

Implementation of some Machine Learning algorithms with source code ...

A java implementation of a neural net. Solving the 8 queens problem with a genetic algorithm (from CS478). Implementation of the k-means algorithm using MDL for model selection. An experiment with evolving artificial neural networks. An implementation of the Gibbs sampling algorithm to detect motifs in sequences (from CS478). Source code of Shotgun aka Ensemble Selection. An implementation of greedy agglomerative clustering with basic graphing. Quick and dirty k-nearest neighboor.