Zerith - - AI-Powered Contract Risk Analysis

Zerith is an AI-powered contract analysis platform designed to make legal documents more understandable, accessible, and practical for everyday users. Modern contracts are often extremely difficult to read due to complex legal language, hidden clauses, automatic renewals, liability traps, and information overload. Zerith addresses this problem by allowing users to upload or paste contracts directly into the platform and receive fast AI-generated analysis in real time. The platform generates simplified summaries, highlights important clauses, and identifies potentially risky sections that users should pay attention to before signing agreements. Zerith was intentionally designed around real-world responsiveness and practical usability rather than relying purely on oversized language models with extremely high latency. From a technical perspective, the system uses a React + Vite frontend combined with a Go backend responsible for document orchestration, chunking, inference communication, and response aggregation. AI inference is powered locally through Ollama and Mistral 7B running on GPU-accelerated infrastructure using ROCm and AMD Developer Cloud resources. A major focus of the project was deployment reliability and inference optimization. The platform includes HTTPS reverse proxy deployment, custom domain routing, chunked large-document handling, concurrent request support, and a smooth frontend analysis workflow optimized for perceived responsiveness. Real-world benchmark testing achieved: * ~4.7 seconds for a 578-word contract * ~5.5 seconds for a 700-word contract * ~9.2 seconds for a 6000-word corporate contract Zerith was built as a complete end-to-end AI system focused on practical engineering, usability, deployment, and inference efficiency rather than simply wrapping prompts around a large language model.