Raoui Ben Aissa

• Co-Built a blended classifier predicting whether an input sentence is a drug, a posology or other • Designed a matcher recognizing input drugs from a list of drugs using multilevel similarity measure • Co-Designed a NER model extracting relevant entities (dose, frequency,..) from an input posology • Successfully orchestrated all the above in a REST API to extract information from prescriptions

✸ 𝐂𝐡𝐮𝐫𝐧 𝐩𝐫𝐨𝐣𝐞𝐜𝐭 : • Improved existing analytical Data Marts quality (missing files, Data reliability, reconciliation) • Co-engineered the right set of features and Built land line models predicting mobile Churn • Automated Data Extraction & cleansing of 4Mln user and up to 30 variables • Predicted correctly 83% of churners & clustered them to identify top reasons causing churn • Connected & Implemented the churn classifier & the clustering model in prod successfully ✸ 𝐁𝐮𝐬𝐢𝐧𝐞𝐬𝐬 𝐦𝐨𝐧𝐢𝐭𝐨𝐫𝐢𝐧𝐠 𝐩𝐫𝐨𝐣𝐞𝐜𝐭 : • Reviewed & evaluated with front office team the current reward system of Orange retail stores • Clustered all of Orange retail stores based on their performance and sales quality & quantity • Created & validated a new metric evaluating every retailer and improved current reward system ✸ 𝐂𝐨𝐯𝐢𝐝𝟏𝟗 𝐩𝐫𝐨𝐣𝐞𝐜𝐭 (𝐢𝐧 𝐚 𝐜𝐨𝐥𝐥𝐚𝐛𝐨𝐫𝐚𝐭𝐢𝐨𝐧 𝐰𝐢𝐭𝐡 𝐭𝐡𝐞 𝐡𝐞𝐚𝐥𝐭𝐡 𝐦𝐢𝐧𝐢𝐬𝐭𝐫𝐲) : • Designed an analytical DWH tracking EMS daily call traffic during country lockdown • Traced high EMS call traffic users & their network and reported to the health observatory • Co-prototyped a base learner predicting covid19 carriers based on telco behavior • Successfully incorporated the base learner with the health ministry biology team classifier

• Followed a special training in quantitative credit risk & credit scoring industry norms • Collaborated with the credit team to engineer a set of features to predict credit scores accurately • Built, improved & tuned a predictive model forecasting credit scores from scratch • Created a validation system for the predictive model relying on an internal credit scoring system

• Documented the commercial department need for a Bank branches performance estimation tool • Co-developed a web application to meet the commercial department need • Increased the Bank branches performance estimation accuracy by nearly 0.125%