SMART INDIA HACKATHON 2025 · TRANSIT INTELLIGENCE

The answer was already in the file.

MetroMind indexes municipal transit records, scanned maintenance blueprints, and safety audits into a grounded vector intelligence system. Built for transit authorities who need verified answers in milliseconds, not manual archive hunts.

<240ms
Query Latency
100%
Page Attribution
Milvus RAG
Local Vector Store
Line 2 · Operational Records
Transit engineering analyst querying MetroMind document intelligence system
What is the scheduled ultrasonic bogie inspection cycle?<240ms
Verified Grounding99.4% Match

Bogie suspension inspections are mandated every 15,000 km or 30 days. Critical wear thresholds require immediate ultrasonic testing.

Maintenance_Report_Q2_2024.pdf · §4.2, p. 18Source Verified
250,000+
Scanned Pages Indexed per Node
<240ms
Average Semantic Lookup Latency
Zero
Synthetic LLM Hallucinations
100%
Air-Gappable On-Premise Execution
INTERACTIVE DOCUMENT RETRIEVAL

Query municipal archives in real time.

Select common transit queries below to see how MetroMind isolates the verified clause, paragraph, and page reference from dense transit archives.

metromind / document-intelligence
Local Milvus Vector Store: 1,480 Docs Indexed
REPOSITORY FOLDERS
VERIFIED CITATIONS
Zero Hallucinations

Answers cite exact page, section, and original PDF blob.

Sample Questions Across Real Transit Documents:
Q
MetroMind Grounded Answer
Operations & Maintenance
According to the Q2 2024 maintenance report, the scheduled track inspection frequency has been increased from quarterly to monthly.
Maintenance report - Section 4.2
Track Integrity Protocols · Page 18 - Q2 2024
Embedding Engine: all-MiniLM-L6-v2 (384-dim)
Retrieval Latency: 38ms (Milvus HNSW)
ENGINEERING CAPABILITIES

From scattered archives
to grounded operational answers.

MetroMind replaces slow manual searches across legacy folders with automated OCR parsing, dense vector embeddings, and strict citation verification.

Multi-Modal OCR & Blueprints

Parses scanned TIFFs, multi-column reports, complex blueprints, and maintenance spreadsheets without destroying spatial layout or table structures.

Tesseract + LayoutLMv3

Zero-Hallucination Grounding

Every answer is anchored to exact PDF page numbers, clause identifiers, and bounding boxes. If a claim cannot be verified in the source file, it is rejected.

Byte-Level PDF Attribution

Departmental RBAC & Audit

Enforces strict role-based access control. Field technicians, auditors, and municipal directors only retrieve documents authorized for their clearance tier.

PostgreSQL Access Policies

Sub-Second Hybrid Search

Combines BM25 lexical keyword matching with Milvus dense vector embeddings to handle exact component model numbers and natural conversational queries.

Milvus HNSW Vector Graph
Elevated metro train viaduct smoothly moving through modern city and forest
Smart India Hackathon 2025 Engineering Showcase

Important information keeps cities moving.

MetroMind was created by Rounak Neema as part of Smart India Hackathon 2025 to help transit authorities, engineers, and municipal agencies find critical operational answers instantly without lost records or manual page searches.

SYSTEM WORKFLOW

How MetroMind operates.

From scanned PDF intake to sub-second grounded answers, the architecture is designed for deterministic attribution and high throughput.

STEP 01

Ingestion & OCR Workers

Documents enter the Go API Gateway and are queued via RabbitMQ. Distributed Python workers run OCR extraction, table segmentation, and text tokenization asynchronously.

STEP 02

Chunking & Milvus Indexing

Extracted text is partitioned into semantic chunks with overlapping boundaries, embedded via all-MiniLM-L6-v2, and indexed into a local Milvus HNSW vector database.

STEP 03

Grounded RAG & Citation

User queries retrieve the top-k most relevant document chunks. The answer synthesizer verifies exact clause references and returns direct page links with zero fabrication.

BUILT FOR TRANSIT OPERATIONS

Where verified records lead to action.

Operations & Track Maintenance

Quickly isolate details from track inspection records, bogie diagnostics, third-rail specifications, and scheduled overhaul logs.

Safety & Regulatory Compliance

Verify compliance with NFPA 130 ventilation rules, CENELEC SIL-4 signaling standards, and audit checklists across municipal divisions.

Station Expansion & Projects

Track contractor deliverables, station expansion milestones, architectural engineering blueprints, and vendor milestone approvals.

Internal Municipal Auditors

Enable cross-functional auditors to search historical archives without relying on personal memory or siloed desktop files.

TECHNICAL SPECIFICATION SUBPAGE

Inspect Distributed Microservices & Milvus Topology

All low-level Go gateway benchmarks, RabbitMQ queue metrics, Python OCR workers, and PostgreSQL RBAC schema details belong strictly on our technical architecture page.

TRANSIT INTELLIGENCE KNOWLEDGE BASE

Frequently Asked Questions

Showing 10 of 10 questions

MetroMind is an enterprise document intelligence and semantic search platform. It ingests complex municipal documents, scanned transit agreements, architectural PDFs, and engineering diagrams, parsing raw images into structured text, tabular data, and vector embeddings for instant semantic lookup.

Interested in MetroMind?
Type your email to connect directly

This system was engineered as a technical showcase. If you are interested in similar builds, architectural consulting, or private deployment, enter your email below.