Samir Ibbou

Summary: Implementation of an intelligent chatbot (in POC) for technical teams, enabling quick and precise searches in internal documentation (Wiki, Jira, office files), to streamline access to technical knowledge as part of the DAFNE platform redesign.

Objectives: Develop a RAG (Retrieval-Augmented Generation) + MCP (Model Context Protocol) solution to query heterogeneous sources (Confluence Wiki, Jira, Office/PDF documents). In sovereign mode (with Ollama/vLLM) vs. AWS Bedrock accessed via Lambda functions. Offer a multilingual search experience (FR/EN) with optimal accuracy through the combined use of embeddings and reranking. Provide a web interface and an internal API to facilitate integration into IT team workflows (Gitlab + Kubernetes).

Technologies, skills=[]

Summary: Construction of the new counterparty risk rating platform (DAFNE) for Crédit Agricole SA. 50 Crédit Agricole entities impacted. Team of 80 people.

Objectives: My role as IT solution manager consisted of: Implementing a scalable, high-performance, and resilient cloud architecture to handle the multiple daily requests from 30,000 users from 50 Crédit Agricole entities. Not to mention all the batch data import and export processes that had to run during the day. All of this had to, of course, comply with the group's standards in terms of architecture, security, and data privacy rules. Recruit/train development teams in Lisbon. Deliver the platform within the allocated time and budget. Reduce user request response times by 50% to 80% through the implemented architecture. Train technical teams on using the new platform. Implement monitoring and proactive incident management to ensure platform availability. Reduce operating and methodology evolution costs by 60% through optimization of resources and processes (on BPM Drools).

Technologies: The most suitable architecture was the CQRS pattern coupled with the Event Sourcing pattern. Write access is performed on a PostgreSQL database where the consistency of the aggregate root (DDD concept) would be ensured. Then, data reading is done on a NoSQL server cache (Redis) where denormalized data (key/value) would be accessible for all types of requests. Data synchronization is ensured by a Kafka message broker. All messages are also saved, so the Redis server cache can be rebuilt at any time., skills=[]

Summary: Building the IT system of an InsurTech with a back-end based on a Microservices and SaaS architecture orchestrated by K8s and a Web/Mobile front-end on React JS and Native. Team of 11 people (5 Back and Front developers in CI/CD, 3 QA/QC, 1 Product Owner, 1 Scrum Master, 1 CTO/Software Architect) and Budget/Deadlines: + €1M over 1 year.

Context/Objectives: As CTO of IT Unchained, we set up a public cloud hardware and software infrastructure for a major client to manage its insurance business, similar to the online insurer "Alan".

The adopted technology is container orchestration (Docker/Kubernetes) on a public cloud (EC2 on AWS) using a very good market platform, Tectonic (CoreOS). For the continuous integration/deployment platform (DevSecOps), we opted for GitLab CIRunner (CI/CD toolchain) with Molecule/TestInfra (Python) for configuration management tests (Ansible), TerraTest (Go) for IaC tests written with Terraform (HashiCorp), and DAST (based on OWASP ZAP Full Scan) for dynamic vulnerability detection.

Cluster/node/pod monitoring/alerting + all application/DB/Mobile logs are handled by Tectonic using Fluentd/Prometheus/ElasticSearch tools combined with visualization on Grafana (dashboard).

For persistence, we chose the distributed NoSQL database DynamoDB, compatible with MongoDB (highly performant, scalable managed AWS storage offering with ACID transactions and schemaless). It's the ideal database for storing JSON documents from third-party SaaS applications used in the solution (e.g., Stripe, SendinBlue, Intercom,...), skills=[{"id": "Amazon DynamoDB", "name": "Amazon DynamoDB"}, {"id": "Terraform", "name": "Terraform"}, {"id": "SendinBlue", "name": "SendinBlue"}, {"id": "Intercom", "name": "Intercom"}, {"id": "GitLab CI Runner", "name": "GitLab CI Runner"}]