About this role
Team: AI Enablement Duration: Approximately three months Location and working arrangement: Brussels HQ Working language: English About the internship Mobilexpense is looking for an AI Engineering Intern to join the AI Enablement team and contribute to a substantial AI project with real internal users and business value. During the internship, you will take ownership of one primary project, from technical exploration and architecture through development, testing, deployment, and demonstration. You will work closely with our AI Engineer and collaborate with teams such as Security, Data, DevOps, IT, Product, and other business functions depending on the assignment. This internship is intended for someone with solid software development foundations and genuine curiosity about artificial intelligence. We do not expect you to already be an experienced AI Engineer, but we are looking for someone who actively experiments with AI, understands that building a reliable AI application involves more than writing prompts, and wants to develop practical expertise in the field. The final project will be selected according to Mobilexpense priorities, the internship period, and your background and interests. It will be scoped to provide meaningful technical ownership throughout the three-month internship. Possible internship projects Internal Knowledge Assistant using RAG Design and build an assistant that allows employees to retrieve reliable information from internal company knowledge. This would be an end-to-end Retrieval-Augmented Generation project rather than a simple chatbot. The work could include: Document collection, parsing, cleaning, and metadata extraction. Chunking strategies for structured and unstructured documents. Generating and managing embeddings . Storing content in a vector database . Implementing semantic vector search , hybrid search, or reranking. Building a complete RAG pipeline using an OpenAI model. Producing grounded answers with source citations. Applying permissions and access boundaries to retrieved content. Creating an evaluation dataset to measure retrieval quality, answer accuracy, and hallucinations. Exposing the solution through an API or a small internal application. Documenting the architecture, limitations, and possible production path. The objective would be to deliver a technically credible prototype and demonstrate, through evaluation, whether it provides sufficiently reliable answers for a defined internal use case. Privacy Aware AI Gateway and MCP Service Design, develop, and deploy an internal service that controls how sensitive information is handled before content is sent to an AI model. The project could include: Building an API or Model Context Protocol server that sits between an internal application and OpenAI. Detecting personal, confidential, or customer-related information. Applying configurable anonymisation, replacement, or filtering rules. Preserving the usefulness and context of the original request after anonymisation. Integrating authentication and user or application permissions. Implementing structured outputs, error handling, audit logs, and traceability. Creating representative test datasets and measuring anonymisation quality. Testing potential prompt injection and data leakage scenarios. Packaging the solution as a deployable service. Collaborating with Security to validate the data-handling approach. Collaborating with DevOps on deployment, configuration, secrets management, monitoring, and CI/CD. Documenting known limitations and operational requirements. The objective would be to move beyond an isolated proof of concept and deliver a service that can be deployed and tested in a realistic internal environment. What you will do Understand a business problem and translate it into a realistic technical scope. Define the project architecture, milestones, and success criteria with your internship supervisor. Research possible approaches and explain their advantages, limitations, and trade-offs. Design and implement an AI-powered application, API, integration, or internal service. Integrate language models with approved data sources, APIs, and business tools. Work with technologies such as RAG , embeddings , vector search , structured outputs , tool calling , agents , or MCP when relevant. Evaluate the solution for accuracy, reliability, latency, cost, security, and privacy. Build tests for both traditional software behaviour and AI-specific behaviour. Use Git and follow normal software development and review practices. Document the architecture, setup, technical decisions, assumptions, and known limitations. Present progress regularly, collect feedback, and improve the solution iteratively. Prepare the solution for deployment in collaboration with the relevant technical teams. Deliver a final demonstration and recommendations for possible next steps.