Verity : EHS AI Compliance Tool

Problem

Why This Project Was Built

EHS teams in the chemical industry juggle thousands of pages of regulations, SOPs, and safety data sheets. Finding the right rule, pre-filling risk assessments, and proving compliance at the site level is slow, manual, and error-prone.

Solution

How It Works

A RAG-backed workspace lets workers search regulations in natural language, auto-generates risk-assessed work permits, issues dynamic checklists, and produces photo-verified PDF proof reports with compliance gap detection.

Core Capabilities

• Regulatory Q&A grounded in 1,214 chunks from 34 ingested PDFs via pgvector + GPT retrieval.
• Work permit workflow with AI risk pre-fill, risk scoring, and supervisor approval chain.
• Dynamic AI-generated inspection checklists tailored to each permit's operation and hazards.
• Photo-evidenced submission flow that emits a signed PDF report and flags compliance gaps.
• Bilingual Korean/English UI with persistent locale preference.

Technology Stack

• Next.js 14 App Router (TypeScript, Tailwind, shadcn/ui)
• FastAPI backend (Python 3.12) with SQLAlchemy async + asyncpg
• PostgreSQL 15 + pgvector for RAG retrieval
• OpenAI gpt-5.4-mini + text-embedding-3-small
• Dockerized multi-service deploy (Postgres + backend + Next.js)

Project Outcomes

• Collapsed multi-hour regulation lookups into seconds of guided Q&A.
• Standardized the chemical-sector permit flow around a single AI-risk-aware workflow.
• Delivered end-to-end proof-of-compliance artifacts without manual PDF assembly.