Raeda AI: 综合健身与营养平台

MicrocosmWorks built Raeda AI's coaching engine using LLM prompt engineering with multiple specialized AI agents: a fitness coach agent that designs workout plans based on user goals and physical capabilities, a nutrition agent that creates meal plans considering dietary restrictions and macronutrient targets, and a wellness agent that coordinates across both domains. Vector database retrieval via Pinecone grounds recommendations in evidence-based nutrition and exercise science.

MicrocosmWorks deployed Raeda AI on Amazon ECS with Fargate and EC2 instances for flexible scaling, with Redis for session caching and response pre-computation. The architecture uses AWS Amplify for the frontend, PostgreSQL for user data, and Pinecone vector database for semantic search across fitness and nutrition knowledge. This combination allows sub-second AI response times even during peak usage while keeping infrastructure costs manageable.

MicrocosmWorks implemented a constraint-based meal planning engine that takes user-specified allergies, dietary preferences (vegan, keto, Mediterranean, etc.), caloric targets, and macronutrient ratios as inputs. The AI agent generates weekly meal plans with grocery lists, substitution options, and preparation instructions, all grounded in nutritional data retrieved from the Pinecone vector database. Plans automatically adjust based on user feedback and progress tracking.

MicrocosmWorks designed Raeda AI's data integration layer to ingest activity data from fitness wearables and health apps, using this real-time data to refine workout intensity recommendations and caloric targets. The Twilio integration enables SMS-based coaching nudges and reminders, while the Node.js backend aggregates data from multiple sources into a unified user health profile that the AI agents reference for personalization.

MicrocosmWorks developed Raeda AI over approximately 14-20 weeks, covering the multi-agent AI system, vector database setup, AWS infrastructure, user-facing frontend, and mobile-responsive interfaces. At MicrocosmWorks rates of $20-$45/hr, a similar AI fitness and nutrition platform typically costs between $25,000-$55,000 depending on the number of AI agents, knowledge base size, and integration requirements.