Applied AI / RAG Engineer (Contract)

About the role

Study Buddy is an AI-powered learning platform for Nigerian students preparing for WAEC examinations, offering personalised explanations, practice questions, mock exams and progress tracking. It is a small founding team based in Abuja. It is hiring an Applied AI / RAG Engineer on a remote contract basis to improve how the AI retrieves trusted educational material, produces useful explanations and proves the quality of its answers. The engineer works directly with the founders on the existing application and grounding pipeline.

Key responsibilities

• Build and improve pipelines for ingesting, cleaning, structuring and indexing approved educational material
• Develop chunking and metadata strategies that preserve subject, topic, source and question context
• Improve retrieval using embeddings, keyword search, hybrid search and reranking where appropriate
• Improve grounded explanations, source citations and handling of questions the material does not cover, using relevant Nigerian examples without unsupported facts
• Measure retrieval relevance, answer correctness, citation accuracy and grounding quality
• Integrate AI features into the TypeScript backend with tests, monitoring and documentation
• Manage latency, token usage, retries and provider costs
• Protect against prompt injection, cross-user data access and sensitive data in logs
• Later phase: evaluate AI-assisted marking against approved rubrics, including uncertainty and human-review handling

Required profile

• Practical experience building and evaluating RAG or other LLM-powered applications
• Ability to distinguish retrieval, content and generation problems
• Experience shipping backend services with tests, version control and production monitoring
• Clear communication, including honesty about limitations and trade-offs

Required skills

• Strong Python and/or TypeScript/Node.js, able to work in an existing TypeScript backend
• Embeddings, vector search, document processing and retrieval evaluation
• Strong SQL and database fundamentals
• LLM APIs, structured outputs and failure handling
• Useful: PostgreSQL vector search (pgvector), Supabase, Prisma, Next.js, Railway, Cloudflare; helpful: EdTech, WAEC formats, human-reviewed evaluation workflows

The first engagement is a fixed milestone of roughly 4-8 weeks, with the option to extend. Candidates apply by email with their CV, availability, expected rate, a RAG project they built and how they measured its quality.