David ELBAZ

Développeur Backend Senior au sein de Prose on Pixels, entité start-up incubée au sein du groupe Havas , spécialisée dans la production créative et publicitaire pilotée par l’IA. Intervention sur Vermeer AI, une plateforme IA créative tout-en-un permettant aux agences, équipes créatives et grandes marques d’orchestrer, sécuriser et monétiser l’usage de multiples modèles génératifs (images, vidéos, 3D, LLM). 🏆 Résultats clés ✔ Transformation d’un MVP IA en une plateforme B2B industrialisable, sécurisée et monétisable ✔ Accélération majeure de l’intégration de nouveaux modèles IA, pour rester compétitif sur le marché GenAI ✔ Mise en place d’un système de crédits robuste, adapté à des APIs IA non déterministes ✔ Plateforme utilisée par ~3 000 utilisateurs (équipes Havas & clients pilotes) ✔ Déploiement progressif à l’échelle du groupe Havas ✔ Monitoring, logs centralisés et alerting temps réel via Grafana 🔧 Contributions techniques ◉ Structuration et refonte de l’architecture backend dans un contexte de startup IA très itérative ◉ Conception d’un système générique d’orchestration inputs/outputs pour intégrer rapidement des modèles IA hétérogènes ◉ Architecture event-driven pour la gestion des crédits (SQS) : blocage temporaire, débit final en cas de succès, ledger propre ◉ Implémentation d’un RBAC fin user × ressource (entités, accounts, workspaces, folders) ◉ Observabilité : dashboards Grafana (performance, erreurs, usage, coûts) et alertes proactives ◉ FinOps IA : optimisation des coûts LLM, arbitrage performance / qualité / coût ◉ Collaboration étroite avec les équipes DevOps, Produit, QA et Créa 🧠 Valeur ajoutée Capacité à industrialiser une plateforme IA complexe dans un environnement de startup incubée, en combinant architecture backend avancée, sécurité, scalabilité et maîtrise des coûts, tout en exploitant l’IA pour accélérer le delivery sans perdre l’ownership ni la qualité du code.

Développeur Full Stack & DevOps lead sur la conception et la réalisation from scratch d’une plateforme RH groupe, visant à uniformiser, sécuriser et gouverner la donnée RH de 18 entités internationales, dans un contexte bancaire fortement réglementé suite à l’intégration de CACEIS au sein du groupe Crédit Agricole. 🏆 Résultats clés ✔ Création du socle central de la donnée RH groupe, garantissant homogénéité, conformité et traçabilité sur 18 pays ✔ Automatisation complète des intégrations mensuelles RH (fichiers de plusieurs milliers de lignes) ✔ Traçabilité groupe ↔ entités : suivi précis des intégrations, retards et non-conformités ✔ Réduction drastique des traitements manuels et des erreurs RH ✔ Alimentation fiable des outils avals (Power BI, ServiceNow, Evidency, MyDev) ✔ Plateforme devenue le point d’entrée de référence pour la donnée RH CACEIS 🔧 Contributions techniques ◉ Pilotage end-to-end du projet en totale autonomie : cadrage métier, architecture, développement, sécurité et déploiement ◉ Conception d’une architecture data from scratch : modélisation SQL performante, historisation, traçabilité et vision long terme ◉ Moteur d’intégration RH : • validations de formats, typages, champs obligatoires et valeurs autorisées • traitements DataFrame vectoriels (Pandas / NumPy) pour performance et scalabilité ◉ Jobs d’intégration asynchrones avec reporting automatique par email ◉ RBAC avancé multi-niveaux (IT groupe, RH groupe, RH entités, collaborateurs) ◉ API REST sécurisée, consommée par le frontend et des services tiers ◉ CI/CD complète : Docker, Jenkins, tests bloquants (unitaires, intégration, non-régression) 🧠 Valeur ajoutée : Capacité à transformer un besoin métier complexe et international en une plateforme critique, robuste et pérenne, combinant vision business, gouvernance data et exigence technique élevée.

Full Stack Developer & DevOps for the MVP of an AI‑driven trading application at TotalEnergies, aimed at automating and securing broker message analysis to optimize energy commodity trading positions AI trading application with real‑time ingestion and price prediction | React • TypeScript • FastAPI (Python) • PostgreSQL • AWS ECS • CloudWatch • AI (NER & forecasting) 🏆 Key Results : ✔ Estimated savings of several hundred thousand dollars per year thanks to the AI‑powered trading application ✔ Ultra‑responsive, AI‑driven interface, with automatic data restructuring and short‑term price predictions ✔ Full customization for each desk: personalized views, color schemes, and filters to trade with one click ✔ Advanced visualization: interactive price history charts and initial price‑movement scenarios generated by the model 🔧 Technical Contributions : ◉ Designed an AI‑driven trading application: real‑time ingestion of IceChat messages and structuring via a fine‑tuned NER model ◉ Data restructuring pipeline: extraction, transformation, and loading into PostgreSQL ◉ Price prediction engine: time‑series forecasting model and FastAPI endpoint delivering instant evolution scenarios ◉ Front‑end with React & TypeScript: real‑time trading table via WebSockets, graphical widgets for history and prediction, annotation form, and user preferences ◉ Backend FastAPI & DevOps: secure ingestion and prediction endpoints, automated CI/CD deployment on AWS ECS, and model/application log monitoring with CloudWatch ◉ Sustainable data architecture: design discussions to ensure data consistency, scalability, and adaptability

Full Stack Developer / Lead Frontend & DevOps on an internal AI-powered platform at Pfizer, built to optimize the evaluation and automation of funding requests as part of the global Grant Program. AI-Powered Grant Management Platform | React • FastAPI • LangChain • AWS • Snowflake • GraphDB • RAG 🏆 Key Achievements: ✔ Cut review time for funding requests from 45 minutes to 45 seconds through intelligent automation ✔ Scaled to 5,000+ funding requests/year with 20+ AI-driven validation rules ✔ Built "Chat with AI at Pfizer", a GPT-based assistant leveraging GraphDB and RAG (Retrieval-Augmented Generation) for contextual and traceable answers ✔ Delivered full CI/CD pipeline using Docker, AWS ECR, and EC2 auto-deployment ✔ Developed a real-time sync pipeline from Oracle DB to Snowflake for live data usage 🔧 Technical Contributions: ◉ Led frontend architecture using React, Vite, and TypeScript, ensuring high performance and smooth integration with AI modules ◉ Built a dynamic dashboard for monitoring and managing funding requests:   • Natural Language Search with LangChain + LLM   • Table views, CSV exports, and automated PowerPoint generation ◉ Developed the “Initial Checks” automation module, drastically reducing manual verification:   • Executes 20+ AI rules and real-time lookups   • Features error logging for diagnostics and audit ◉ Created “Chat with AI at Pfizer”: a GPT-based internal assistant trained on Pfizer’s grant documentation, using a GraphDB knowledge base and RAG pipeline to deliver relevant, cited responses ◉ Set up full CI/CD pipeline:   • Docker image management via AWS ECR   • Auto-deployments to EC2 instances ◉ Developed a data sync pipeline from Oracle DB to Snowflake, enabling real-time data refresh

Full Stack Developer at BNP Paribas — Designed automation systems, led the refactoring of core APIs, and developed internal tools used by over 300k employees worldwide. 🏆 Key Achievements: ✔ Complete refactoring of the internal marketplace API into a scalable microservices architecture, enabling better modularization, independent deployments, and increased service resilience ✔ 120+ automated requests/day, significantly reducing support team workload ✔ Tools and APIs actively serve 300k+ employees globally, improving IT efficiency and service delivery ✔ Successful cloud migration boosting performance, scalability, and security ✔ Delivered robust and maintainable solutions in line with BNP’s enterprise-level security standards 🔧 Technical Contributions: ◉ Refactored the core internal marketplace API (Node.js/TypeScript) into a microservices-based architecture, allowing domain separation, scalability, and independent module updates ◉ Developed and maintained automation solutions using a Go-based orchestrator and a Node.js/TypeScript backend API ◉ Built a log tracking and email release bot for messages quarantined by Cisco IronPorts, replacing manual processing → 💡 Impact: 120+ daily requests handled automatically, freeing up two full-time support agents ◉ Designed a flexible MongoDB-based architecture to run dynamic backend scripts, enabling real-time API integrations, secure transformations, and filtering pipelines ◉ Led the migration of the backend API from physical servers to the IBM Cloud, using Docker, Kubernetes, and HashVault for secret management ◉ Collaborated closely with Vue.js front-end teams to ensure seamless integration and optimal UX (JIRA → GitLab migration) ◉ Promoted development best practices: ESLint, Test-Driven Development (TDD), Pair Programming, Code Reviews ◉ Built a custom UI for Product Owners to generate tailored API payloads for each product (React, TypeScript, ViteJS) → 📦 Deployed via GitLab CI on BNP’s internal cloud infrastructure

Frontend Developer on an ERP designed for landlords managing multiple rental properties in the short-term rental market - +25% increase in average bookings for clients. 🎯 Goal: Centralize booking management, calendars, finances, and OTA (Airbnb, Booking, Expedia...) synchronization in a modern ERP solution. ◉ Stack: React + Ant Design Pro (Frontend), PHP Symfony (Backend), MySQL ◉ Full integration with Channel Managers and major OTA platforms ◉ Developed interactive calendars, dashboard, accounting features, and dynamic invoice generation ◉ Worked using Agile methodology: sprints, daily stand-ups, backlog refinement ◉ Conducted functional analysis in close collaboration with beta clients involved throughout the development process ◉ Maintained continuous feedback loops with users to optimize the product and UX ◉ Impact: +25% increase in average bookings for clients after implementation of the ERP

Full Stack development mission aimed at providing Product and Management teams with an interface to easily edit the visuals and forms sent to each patient. ◉ React as the main framework for client-side development ◉ Firebase implemented for authentication and RealTime Database ◉ Designed an intuitive UI/UX to allow non-technical users to edit questions stored in the database ◉ Collaborated in a mixed team of Juniors and Seniors (UI/UX Designer, Developers, Project Manager, Mentor) ◉ Successfully delivered all milestones within the deadlines defined by the Lead Project Manager

Development of a music application: ◉ Stack: React (Material UI, Redux, Custom Hooks, Sass) & Node.js (Express.js) & MongoDB ◉ PWA ◉ WebSockets (Live Updates) ◉ Google APIs ◉ Playlist creation and editing ◉ User management ◉ Drag & Drop, and more...