GPT-4o fine-tuning for low-resource spoken languages: Dimagi fine-tuned gpt-4o-mini for machine translation of English to Sheng (Swahili-English slang) and gpt-4o for English to Chichewa to deliver health education chatbots for use in Kenya and Malawi achieving 250% and 16% BLEU improvement, respectively. This session will cover the challenges of delivering LLM-based health education tools in low-resource languages to previously unreachable populations, the modular architecture enabling future adaptation to additional languages, our approach to sourcing additional data from professional translators and open-source datasets and fine-tuning OpenAI models for machine translation.