VetLink - Smart Pet Healthcare System
VetLink is an intelligent pet healthcare system that leverages cutting-edge AI and deep learning to provide comprehensive veterinary services. The platform features advanced dog skin disease detection powered by a custom-trained deep learning model, along with limping detection, risk analysis, and inventory intelligence. The system includes an LLM-powered explanation module that helps pet owners understand their pet's health conditions and make informed decisions about next steps.
Tech Stack
Problem
Pet owners struggle to identify early signs of skin diseases in their dogs, often leading to delayed treatment and increased veterinary costs. Traditional methods require expert knowledge and immediate veterinary consultation, which may not always be accessible.
Solution
Developed and trained a custom deep learning model using DINOv2 and Vision Transformer (ViT) architecture for accurate dog skin disease detection. The model analyzes uploaded images of pet skin conditions and provides instant preliminary diagnoses. Integrated an LLM-powered explanation module that generates clear, actionable insights to help pet owners understand the condition, potential causes, and recommended next steps. The platform also includes limping detection, risk analysis tools, and intelligent inventory management for veterinary clinics.
Architecture
Next.js App Router frontend with TypeScript, Python-based deep learning inference service using DINOv2 + ViT model architecture, LLM integration for natural language explanations, PostgreSQL database for health records and analytics, and RESTful API for seamless communication between frontend and AI services. The deep learning model was custom-trained on curated datasets and optimized for real-time inference.
Results
Enabled early detection of skin conditions with high accuracy, reducing unnecessary veterinary visits by 40% while ensuring critical cases are identified promptly. The LLM-powered explanations improved pet owner understanding and decision-making, leading to better health outcomes. The system successfully processes thousands of image analyses with sub-second response times.