We are looking for an AI Engineering Intern to build a sovereign AI assistant that lets organizations search, understand and summarize their internal knowledge securely, with all models and data kept within their own infrastructure. The goal is to deliver a trusted, bilingual assistant for government and enterprise teams.
The project involves building a solution capable of:
- Answering questions in Arabic and English, tuned to institutional terminology.
- Retrieving information across policies, circulars and archives with a hybrid RAG pipeline (semantic + keyword search, re-ranking).
- Summarizing documents and extracting key insights, with citations to the source.
- Running fully on-premise with self-hosted, optimized LLMs.
- Enforcing encryption, role-based access control and complete audit trails for every query and document interaction.
This internship provides hands-on experience in LLM engineering, retrieval systems and secure AI deployment.
Technical Skills
- Solid Python and experience with LLM frameworks (LangChain or LlamaIndex).
- Understanding of RAG, embeddings and vector databases.
- Experience building REST APIs (FastAPI or similar).
- Basic knowledge of Docker and Linux.
Bonus Skills (Not Required but Appreciated)
- Experience with Arabic NLP or multilingual models.
- Knowledge of LLM serving and quantization (vLLM, GPTQ, AWQ).
- Familiarity with LLM evaluation and guardrails (RAGAS, Llama Guard).
Tech Stack
- LLM & RAG: LangChain or LlamaIndex, vLLM, open-source LLMs
- Search: Qdrant or pgvector, BM25, re-rankers
- Backend: Python, FastAPI, PostgreSQL
- Security: Keycloak (RBAC), encryption, audit logging
- DevOps: Docker, Git, on-premise GPU servers
