Akram C.

AI & Automation Team - Data CoE (Arval IT department) GenAI / RAG Assistant (500+ Users) - Built the initial internal assistant POC across the RAG pipeline, vector search stack, and user interface. Extended from POC to production. Also contributed to product analytics and the monitoring layer. Email Classification (NLP, 2 countries) - Led deployment of a production email-classification system across 20+ mailboxes in 2 countries. Refactored legacy code into a modular Python service with async orchestration (Redis + Celery), exposed via FastAPI, and deployed on Kubernetes with Helm. Agentic AI Platform (MVP) - Independently built a multi-agent orchestration platform connecting Jira, Confluence, and Bitbucket via LangGraph + MCP. Added a hybrid knowledge base (ChromaDB + Neo4j + Mem0) and observability with Arize Phoenix + OpenTelemetry. Built for enterprise engineering workflows, with EU AI Act and GDPR constraints in mind. MLOps: Enterprise Architecture (Professional Thesis) - Designing an end-to-end MLOps architecture bridging DS (Domino/MLflow) and IT (Kubernetes/GitLab CI/ArgoCD). Following the Google MLOps maturity model to move the organization from manual handoffs (L0) to automated, CI/CD-driven model deployment (L2). Defined the deployment contract between DS and IT. Architecture grounded in 10+ research papers. Includes Tanzu-to-ROKS migration planning (Red Hat OpenShift Kubernetes Service). Data Platforms & ETL - Built Python ETL pipelines, internal data-quality tools, Django admin app, and Metabase functional dashboards.

DigiLab / MyArval teams - DCX Tribe (OneIT department) - Built the first internal GenAI assistant POC (LangChain + Chainlit) for RAG over internal documentation, with a multi-assistant setup and dynamic LLM selection. The POC was later validated and extended into a production system. - Set up Apache Superset and created initial dashboards to monitor GenAI assistant usage, along with other functional dashboards. - Built multiple Python ETL packages to support internal data workflows.