Accelerating MSE Compliance using AI (RAG System)
The BIS Standards Recommendation Engine is an AI-powered system designed to help Micro and Small Enterprises (MSEs) quickly discover relevant standards published by the Bureau of Indian Standards.
Instead of manually searching through lengthy documents, users can simply enter a product or material (e.g., cement, steel, AAC blocks), and the system will:
MSEs often struggle with:
Complex BIS documentation Time-consuming manual search Lack of technical understanding Risk of non-compliance
This project automates the process using Retrieval-Augmented Generation (RAG).
This system combines:
Information Retrieval (TF-IDF) → Finds relevant standards AI/LLM (optional) → Generates explanations Flask Web App → User-friendly interface
✔ Fast keyword-based search (milliseconds) ✔ No GPU or heavy ML required ✔ Clean web UI for easy interaction ✔ Explanation for each recommendation ✔ Works offline (no API required) ✔ Supports real BIS PDF ingestion
bis-rag-engine/
├── inference.py ← Judge entry-point (required)
├── eval_script.py ← Evaluation metrics (required)
├── requirements.txt
├── README.md
├── src/
│ ├── rag_pipeline.py ← Core RAG logic + data source switching
│ ├── ingest.py ← BIS SP-21 PDF parser & chunker
│ └── app.py ← Flask web app with source indicator
└── data/
├── BIS_SP21.pdf ← Place official PDF here (not committed)
├── chunks.json ← Generated by ingest.py (auto-loaded)
├── public_test_set.json ← 10 sample queries
├── ground_truth.json ← Expected standards for public set
└── results.json ← Generated by inference.py
Step 1: User Input
User enters a query (e.g., cement)
Step 2: Retrieval TF-IDF algorithm finds matching standards Keyword boost improves accuracy
Step 3: Ranking Top 3–5 relevant standards are selected.
Step 4: Explanation AI or rule-based logic explains why each standard applies.
Deploy:
https://bis-rag-engine.onrender.com