An explainable water-risk platform for visarjan ghats: historical analysis, ghat prioritisation, data-quality checks, sampling plans and reports in one place.
The problem
Visarjan activities can create environmental concerns around water bodies, while historical water-quality observations may be scattered across records and difficult for municipal officers to interpret quickly. Existing data may also contain missing measurements, inconsistent site names, and limited traceability, making it challenging to identify where further monitoring is needed.
GhatShield AI addresses this gap by transforming historical water-quality data from Bhopal’s visarjan ghats into an explainable decision-support workflow. It helps identify potential water-quality concerns, highlight missing evidence, compare historical trends, and prioritise locations for additional sampling. The goal is to support more transparent, data-driven municipal decisions without treating missing data as evidence of safety.
How it works
GhatShield AI is an explainable water-risk and resource intelligence platform designed for municipal decision-making around visarjan ghats. It brings historical water-quality analysis, ghat prioritisation, data-quality monitoring, sampling resource planning, evidence-grounded AI assistance, and report generation into one workflow.
Its key innovation is connecting risk priorities with supporting observations, reference thresholds, source records, and data-quality limitations. Instead of merely displaying charts, the platform helps officers understand why a location may require attention and what information should be collected next. Its missing-data intelligence prevents unavailable measurements from being interpreted as zero or as proof of safety. The sampling optimiser supports configurable planning, while the AI Action Copilot helps translate technical findings into understandable explanations and structured reports.
The prototype is designed around historical observations from 2023–2025 and supports evidence-based prioritisation rather than making live monitoring or official water-safety claims.
How Claude was used
We used Claude as an AI-assisted development partner to structure the project architecture, plan backend modules, design API contracts, generate implementation guidance, troubleshoot errors, and refine the product workflow. We used it to reason through historical data validation, missing-value handling, explainable risk prioritisation, sampling optimisation, and evidence-grounded reporting.
Claude also helped us organise the implementation into manageable modules and develop prompts for the user interface and backend. Our approach keeps core calculations and data validation separate from generative AI so that recommendations can be tested against the underlying evidence. Claude-assisted outputs are reviewed and tested before being treated as part of the working prototype.
Team
Team Setu