AI/NLP Engineer — FarmConnect Advisory Engine

About the role

This is the most technically challenging role on FarmConnect and the one with the greatest potential for impact. You will build an AI advisory engine that can answer a Kenyan farmer's question — in Swahili, Kikuyu, or English — about the best time to plant, why their maize leaves are yellowing, or what market price they should accept for their tomatoes. The engine will combine a fine-tuned LLM with a retrieval-augmented generation pipeline that draws on an agricultural knowledge base of over fifty thousand documents. You will: curate and preprocess the training corpus (Kenya agricultural extension guides, CIMMYT publications, County agricultural reports, farmer outcome data); fine-tune a multilingual LLM (likely Gemini or LLaMA with multilingual capability) on this corpus; build the RAG pipeline using a vector database for knowledge retrieval; implement the WhatsApp Business API integration that routes incoming farmer messages to the advisory engine and delivers responses; build the fallback escalation system that routes complex queries to human agronomists; and implement the answer quality evaluation framework with agricultural domain expert review. You will measure model performance on a held-out evaluation set of real farmer questions with expert-verified answers, targeting above eighty percent answer accuracy on the primary crop types. You will monitor response quality in production and implement human-in-the-loop correction workflows. Strong candidates have experience with LLM fine-tuning or RAG pipelines, NLP in low-resource African languages, and building conversational AI for users with limited literacy. This role includes co-authorship on any academic publications from the research.