Back to Ideas
Defence Innovation
2025

DRISHTI — AI-Based Design Change Impact Analysis

AI system for design change impact analysis and cost estimation for Mazagon Dock Shipbuilders

Neo4jMistral-7BReactD3.jsGraph-DBAIPython

The Problem

When a design change is proposed for a naval vessel, understanding its cascading impact across interconnected subsystems — structural, electrical, piping, ventilation — and estimating the resulting cost is currently a slow, expert-dependent process. DRISHTI MDL #70 asks for a system to automate and accelerate this.

Proposed Architecture

Neo4j Graph Database — Models the relationships between ship components, subsystems, and design parameters. Impact analysis is fundamentally a graph traversal problem, and a graph database makes this explicit rather than forcing relational joins.

Fine-tuned Mistral 7B — Handles natural language understanding of engineering change requests and generates cost estimates grounded in historical data. Open-source model chosen for deployability in defence contexts.

React + D3.js Dashboard — Visualizes impact chains and cost breakdowns, giving engineers an interactive view of how a proposed change ripples through the vessel's design.

The Core Differentiator

The same config-driven similarity matching approach I built as a production microservice at Infosys powers this system's ability to identify which components are affected by a proposed change. Production-tested matching logic, repurposed for a defence application. It's not a pivot — it's a straight line.

Roadmap

14 to 16 months from proposal to deployment. Currently in the active proposal stage for the DRISHTI defence innovation challenge.