ATHENIS is a web platform built during HackOdisha 5.0 to simplify complex legal documents and surface risks in plain language. The goal was to make contracts, policies, and legal text more accessible to individuals and small businesses without legal expertise. The system processes long legal documents in real time and streams simplified output back to the user as analysis progresses. It is deployed on decentralized GPU infrastructure to keep inference fast and cost efficient. What the system does: - Accepts legal documents in PDF, DOCX, or text formats - Simplifies clauses into plain language while preserving meaning - Flags potentially risky clauses using AI assisted analysis - Streams results live to the UI for long documents - Supports multilingual input and long form text processing My contributions: - Contributed to backend services handling document ingestion and AI inference workflows - Designed and built the frontend interface for document upload, live analysis view, and result visualization - Integrated frontend and backend using real time Server Sent Events for streaming output - Containerized services using Docker and prepared the application for decentralized deployment - Managed deployment on the Akash Network with GPU enabled inference - Set up CI and CD pipelines for automated builds and testing Technical focus: A key challenge was handling long legal documents without blocking the UI. This was solved using streaming responses and careful backend orchestration, allowing users to see partial results while processing continues. Tech stack: Python, Flask, PyTorch, Transformers HTML, CSS, JavaScript, Tailwind CSS Docker, Akash Network, CI CD pipelines Server Sent Events for real time updates Outcome: Top 6 finalist at HackOdisha 5.0 in the Akash Network track, selected from a large national level participant pool. GitHub: https://github.com/gabsgj/athenis Frontend Deployment: https://athenis.gabrieljames.me/