Contributing multilingual support to Microsoft Bot Framework v4 so chatbots can serve any language through a translation middleware over LUIS and QnA Maker.
Our IEEE paper on gender-aware Arabic-to-English spoken language translation, injecting speaker and listener gender tokens into NMT to preserve agreement and reduce bias.
Competing in AI Challenger's English-to-Chinese translation contest with an attentional seq2seq NMT model and BPE subwords, scoring 25.5 BLEU on the test set.
Our Microsoft Research paper on generating synthetic parallel data for low-resource neural machine translation using word embeddings and three-way semantic projection.
Tackling Google's TensorFlow Speech Commands Kaggle competition with MFCC, mel and log-mel features, VGG and ResNet keyword spotters, and audio data augmentation.
Two years at Valeo Automotive: software development across MATLAB, Simulink, C# and Unity, building an internal wiki and an automated unit-test generation tool.
A MATLAB library I built at Valeo to automate Simulink model generation, exposing a simple scripting API on top of Embedded Coder for C code generation.
Classifying human actions in Pascal VOC 2010 still images using dense SIFT features, PCA and random forests in MATLAB, the project that pulled me into machine learning.
My Bkam internship as a software engineer: shipping a Windows Phone price-comparison app and building Python web crawlers with product name clustering.
