Chizzy-codes/Kenya_Clinical_Reasoning_Project_Extended

My solution to the Kenya Clinical Reasoning Hackathon on Zindi # Kenya_Clinical_Reasoning_Project_Extended My solution to the Kenya Clinical Reasoning Hackathon on Zindi # Can your model match real clinicians in rural Kenyan healthcare? In many parts of the world, frontline healthcare workers make life-or-death decisions under pressure, with limited resources and specialist support. This challenge takes you to the heart of Kenyan healthcare, where nurses working across diverse counties and health facility levels face real-world clinical cases every day. In this challenge, you'll be given 400 authentic clinical prompts—each one a carefully crafted vignette combining a nurse’s background and a complex medical situation. Your task is to predict the clinician’s response to each scenario, replicating the reasoning of trained professionals as closely as possible. The vignettes span a wide range of medical domains, from maternal and child health to critical care, and were originally evaluated by expert clinicians and leading AI models (including GPT-4.0, LLAMA, and GEMINI). Each prompt includes details like the patient's presentation, nurse experience level, and facility type, simulating the nuance and challenge of real clinical environments in Kenya. This dataset is small—only 400 training and 100 test samples—but that’s because collecting high-quality, expert-labelled medical data is hard. These are real-world cases, and every entry reflects the constraints and pressures faced by healthcare workers in underserved regions. In resource-limited settings, clinical decisions must be fast, accurate, and sensitive to both patient condition and system limitations. Your solution should aim to reflect that balance.