• Worked on Non-Second price bidder (NPSB) which is a real-time bidding model to maximize the expected profit margin (log-LR, binary search)• Rebuilt all the infrastructure of the bidding team by using Hadoop, Spark, Mesos, and Cuttle scheduler

* Large scale data pipeline- Periodicaly extract data from site server to S3 as the source of batch processing. - Take S3 data for spark jobs, and sink to redshift for effecient analytics and elasticsearch for indexing and searching.- Keywords: spark, redshift, elasticsearch, aws s3* Classified ads streaming service- Import classifed ads in a streaming way via kafka-rest- Create Spark streaming job transforms ads in kafka into a format suitable for the site- Keywords: spark streaming, kafka* Machine Learning: crawler detection- Detecte crawler which pirates site contenu and occupies too much server resources- Model training based on server log batch for a period of time- Online prediction and alert based on server logs stream- Start with logistic regression with the help of blacklist and whitelist- Unsupervised learning, like KMeans gives better results on our dataset- Keywords: spark MLlib, kafka* Initiate Deep Learning: ad2vec- An experimental deeplearning project of data-science lab at leboncoin- Convert each classified ad into a vector (inspired by word2vec) for further usage. (clustering, recommendation, etc)- Keywords: word2vec, NLP* DevOps- Help to make the projects mentioned above running in production in AWS- Scheduling with Airflow- Provisioning with Chef and Packer- Monitoring with DataDog- Reinforce code style convention

- Develop a platform built on the top of Spark framework.- Design, implement & evaluate Machine Learning algorithm with Spark- AWS: EC2, S3, CloudFront, IAM, ElasticBeanStalk- Analytics: SnowPlow, Google Tag Manager- Hadoop eco-system (HDFS, MapReduce, HBase, Hive)- Programming in Scala

- GUI for a wielding generator- QT framework- Multi-thread programming- Programming in C++