Page 88 - Ebook-Book JCMS 2025

Designing an AI Co-Pilot Framwork by Embedding Expert
Logic for Nutritionists in Digital Therapeutics: A Case Study
from Fitsloth
Wuttikorn Ponwitayarat1, Phairot Autthasan1, Nattadhanai Rajatanavin2, Titiporn Tuangratananon2
,
Theerawit Wilaiprasitporn1*
1 Vidyasirimedhi Institute of Science and Technology
2 FitSloth
*Corresponding Author E-mail: theerawit.w@vistec.ac.th
Abstract
This study introduces a comprehensive framework for developing an AI-driven Large Language Model (LLM) co-pilot
system, tailored specifically to assist nutritionists managing obesity within Fitsloth’s digital therapeutics platform. The
primary goal was to embed nutritionists’ detailed decision-making logic and clinical reasoning into the AI system,
ensuring alignment with professional nutritional guidelines. This logic encompasses precise caloric intake management,
macronutrient distribution (carbohydrates, proteins, fats), meal-specific recommendations, identification of frequent
dietary behaviors, and daily meal plan suggestions derived from actual patient food records.
The implemented co-pilot system significantly streamlined nutritionists’ workflows, reducing routine analytical tasks by
approximately 60% and allowing more personalized patient interactions. User feedback highlighted notable strengths,
including concise data summarization, identifying problematic dietary choices, understanding patient nutritional
patterns, and generating practical meal plans.
Future enhancements recommended include long-term dietary trend analysis, comparative monthly nutritional
summaries, and selection of optimal daily meal patterns meeting dietary targets. This structured and adaptable framework
provides significant potential for broader applications across various professional contexts requiring AI-supported
decision-making.
86 Joint Conference in Medical Sciences 2025