Xavier Mortelette

Context:

Scope: Design and development of a dynamic AI agent architecture to improve development team productivity and industrialize the use of LLM models in companies.

Challenges: Improve development team productivity and industrialize the use of LLM models in companies.

Description:

● Design of an AI agent architecture based on LLM models (LLaMA type) with multi-model orchestration.

● Development of a modular agent framework integrating memory management via vector database (PostgreSQL + vectorization), historical logging and reuse of conversational context, as well as orchestration of 'personalities' (agents specialized by use case).

● Implementation of an advanced prompt engineering system (system prompt matrix according to roles and capabilities).

● Design of a middleware layer (intelligent proxy) enabling LLM abstraction, multi-backend call management (OpenAI, local models, etc.), and inference cost and context optimization.

● Development of AI-oriented tools (APIs) with optimized design (structured inputs/outputs, context limitation, data filtering).

● Implementation of optimization strategies: LLM context reduction, task splitting between specialized agents, fine-grained tool management to avoid memory overload.

● Contribution to infrastructure and sovereignty thinking (local model deployment, hardware optimization).

● Execution of business-oriented PoCs (e.g., developer productivity, banking use cases like fraud, analysis, automation).

Technical Skills:

PostgreSQL, APIs, Large Language Model (LLM), ChatGPT, FastAPI, RAG

Context:

Intervention within the Oracle Middleware team at BNP Paribas (BP2I), composed of about fifty people in Paris.

Scope: Elaboration of As-a-Service architectures for middleware offerings, particularly the analysis of integration and migration of Oracle offerings to ExaCC.

Challenges: Ensure compatibility of new cloud architectures with existing systems, secure the migration of critical infrastructures, and optimize the integration of As-a-Service solutions.

Description:

● Analysis of technical and network infrastructures.

● Verification of interconnection and interoperability possibilities with existing systems.

● Preparation of infrastructure and database migration plans.

● Creation of HLD and LLD architecture documents.

● Adaptation of Managed Cloud ExaCC in private datacenters.

● Industrialization of Cloud offerings on internal offerings

Results:

● Improvement of innovation through the migration of Oracle offerings to ExaCC, enabling infrastructure modernization.

● Reduction of operational costs through optimization of As-a-Service architectures.

● Improvement of system reliability through better interoperability and interconnection with existing infrastructures.

Technical Skills:

Ansible, Python, Terraform, Oracle, High-Level Design (HLD), Low-Level Design (LLD)

Context:

Scope: Assistance and consulting for the IT department of the École Nationale d'Architecture Paris Malaquais.

Challenges: Overhaul and modernization of network infrastructures and the information system.

Description:

● Assistance and consulting for the IT department to optimize infrastructures and processes.

● Complete network overhaul, including security, hardware, and architecture.

● Migration of the information system to Office 365.

Results:

● Improvement of network security through a modernized and secure architecture.

● Reduction of operational costs related to network maintenance through an optimized infrastructure.

● Improvement of user collaboration and productivity through the migration to Office 365, enabling better integration of collaborative tools.

Technical Skills:

Linux, CentOS, Hyper-V, Windows, Cisco Ne