Ilan Benichou

En tant que CTPO, mon rôle couvre à la fois les aspects techniques, data et produit. Je dirige le développement de notre application avec une équipe de 5 personnes. Les principales technologies utilisées sont : - Backend : NodeJS, Express et MongoDB - Frontend : React, TypeScript, RTK Query, Chakra UI - Serverless : quelques serveurs en Go et Python (serveur de notifications, webhook) Une partie essentielle de mon travail consiste à garantir des performances backend optimales, en mettant en place des microservices avec Cloud Functions et Pub/Sub, l’utilisation de Cloud Run pour des applications plus complexes. J'ai également travaillé sur l'intégration de partenaires via leurs API (mise en place de Webhook, realTime). Pour la partie Data, j'ai mis en place des processus d'intégration avec Fivetran, créé des couches d'extraction et de transformation avec DBT, et utilisé Looker Studio pour concevoir des tableaux de bord pertinents. En ce qui concerne les processus internes, j'ai favorisé une communication fluide en utilisant Slack et Notion. Je gère des cycles de développement agiles (Sprints), nous documentons largement nos avancées techniques et réalisons des revues de PR sur GitHub afin de maintenir la qualité du code. J'ai également créé la première fonctionnalité B2C de Qeeps : le pass certifié, avec récupération automatique des revenus via des sources officielles. Qeeps est une application immobilière permettant aux locataires potentiels de constituer leur dossier de location. Résultat : un taux de conversion de 5 % et plus de 10 000 nouveaux utilisateurs par mois. En tant que CTPO, je joue un rôle central en reliant les aspects techniques et data, en optimisant les performances et en favorisant une collaboration efficace au sein de Qeeps.

I joined Flink after the Cajoo acquisition as a Lead Data Scientist to work on the Last Mile projects. Design the forecast demand to create planning of riders and pickers in Hub. Forecast demand is related to a lot of project (e.g. order forecast, no-show forecast, productivity of our riders and pickers, …) - Order forecat: We are using different machine learning models to predict number of orders in next 4 weeks. - No-show forecast: We are working on forecasting the percentage of absences in order to adjust the number of riders and pickers.

In charge of Data Science and Data Engineering for the Delivery Team. Work hand-in-hand with Business and Software Engineering Teams. Proactively design, built and maintain robust data models for Delivery Operations stakeholders in dbt. Design the order forecast and staffing web application to automatically create the planning in Hub. - Order forecast: Machine Learning to predict the number of orders for the next 2 weeks. (accuracy ~80%). - Use Streamlit to create web application in Python. - Design an optimisation algorithm to transform the expected number of orders per hour into shifts with the required number of riders. - Increase the productivity of our riders by 15 %. Design the assignation scoring with Software Engineering team to find the best rider. - Optimize scoring to achieve the highest percentage of orders delivered within 15 minutes. - Take into account the delivery costs to have a profitable business model. Design cost engine to compute in real time the cost per order for the preparation and the delivery part. - The aim of this project was to monitor the cost per order target to achieve profitability.

Measurement of Digital and TV campaigns: - Impact of TV advertising campaigns on in-store and e-commerce sales - Impact of digital advertising campaigns on in-store and e-commerce sales Media analysis: Reporting and Calculation of Media Campaign metrics (ROAS, Post-View Conversions, Post-Click Conversions) Recommandations system: - Creation of a product assortment system for food shops. - Choice of K among N products. - Objective: to have the best complementarity between the products and maximize the yield. Google Media: - Working on Google Architecture - Creating Audiences in Google Ads - Computing Media KPIs in ADH

• Churn detection with Machine Learning in Python • Data Visualization with Looker • Creation of Data Warehouse with Redshift(NoSQL) In a SaaS business like Contentsquare customer health and retention are two very important indicators. My mission during my internship was to create a health score for each of our customers. The goal of the project was to make predictions to identify risky customers.

• Development on trading automate (C++, SQL, Python) • Creation of non-regression system for trading automate (C++) Replay of trading day in the past • Analysis of Trading data in Python

• NLP in Python NLTK, Gensim (Word2Vec), Kmeans • Machine Learning and Clustering in Python • NoSQL database management with Neo4j ShorTouch is a technology developed to facilitate relationships between people. It is based on a cross-platform data-set (Facebook, Gmail, Outlook ..) and uses Machine Learning to create an overview of each user and its networks. During my internship I took care of the email analysis part. The goal was to refine the links between people in the same network. To do this NLP processing, I used different machine learning libraries in Python such as NLTK, Gensim, Sklearn NLTK to clean the mails Gensim to use the Word2Vec template Scikit-Learn for clustering with Kmeans