CoachMO – AI Question Generation System

PythonLangChainOpenAI GPT-3.5DeepEval FrameworkJSON Data ProcessingPrompt Engineering Techniques

CoachMO is an AI-driven system designed to generate context-aware questions for endurance training programs. The platform leverages structured workout datasets and Large Language Models (LLMs) to automatically generate high-quality coaching questions tailored to specific training sessions. The system utilizes advanced prompt engineering techniques and evaluation frameworks to ensure that generated questions maintain accuracy, relevance, and logical coherence.

The Problem & Solution

Problem

Traditional endurance coaching platforms rely heavily on manually created question sets or static workout descriptions. This approach introduces several limitations: • Manual question generation requires significant time and effort • Static questions fail to adapt to diverse workout structures • Lack of contextual questioning reduces athlete engagement Existing systems such as Mottiv achieved only about 75% question relevance accuracy, leaving room for improvement in automated coaching systems.

Solution

The system generates personalized workout questions using structured workout attributes such as duration, intensity, nutrition guidelines, and workout structure. A hybrid prompt engineering technique combining Few-Shot prompting and Chain-of-Thought reasoning was implemented to generate contextually accurate and logically structured questions. Generated questions are automatically evaluated using a multi-metric evaluation framework before being stored in a database for real-time use in coaching applications.

Architecture

The architecture includes the following components: • Data Input Layer: Structured workout data stored in JSON format. • Prompt Engineering Layer: Few-Shot examples and Chain-of-Thought reasoning guide the LLM to produce contextual questions. • LLM Processing Layer: OpenAI GPT-3.5 generates questions based on workout attributes. • Evaluation Layer: DeepEval framework evaluates generated questions using multiple metrics. • Storage Layer: Validated questions are stored in a database and served through APIs.

Key Features

Automated question generation using LLMs
Few-Shot + Chain-of-Thought prompt engineering
Context-aware question creation using workout attributes
Automated quality evaluation using DeepEval
Scalable architecture for large training datasets

Key Impact

1
Achieved over 85% accuracy in generated questions, outperforming existing systems
2
Reduced manual question creation effort
3
Enabled scalable AI-driven coaching interactions
4
Improved athlete engagement through contextual coaching questions

Interested in discussing this architecture?

Get in touch