AI Agent·2026

PardaFaash AI

Investigative AI for Indian Public Data

Official Gov

Data Sources

Severity 1-5

Scoring

Deterministic

Pipeline

Overview

PardaFaash AI is an evidence-first, investigative AI agent built to help journalists, researchers, and citizens trace political finance, audit filings, and candidate records in India. Instead of generating vague conversational responses, PardaFaash acts as an automated "evidence desk" — cross-referencing public government datasets to surface financial anomalies, ghost beneficiaries, and compliance gaps.

The agent uses a multi-stage, deterministic pipeline: Query Framing, Parallel Discovery (via Jina API and data.gov.in), Data Grounding, and Fraud & Anomaly Scoring (using a custom Severity 1-5 Framework). It structures the synthesis with narrative summaries, comparative tables, and interactive, strictly sanitized source citations.

Key Features

01Multi-stage, deterministic pipeline designed for investigative research

02Parallel Discovery across data.gov.in, ADR, MyNeta, ECI, and NCRB

03Custom Severity 1-5 Anomaly Framework for fraud scoring

04LLM grounding that strictly prevents hallucination

05Interactive, strictly sanitized source citation chips

Tech Stack

Next.jsReactTypeScriptTailwind CSSHono.jsPostgreSQLDrizzle ORMOpenRouterJina Search API

Project Info

CategoryAI Agent

Year2026

Statusactive

VisibilityPublic

Interested in a similar project?

Get in touch

AgentX · PardaFaash AI

Ask anything about this build

Hey! I'm AgentX — Jay's AI representative. Ask me anything about how I built PardaFaash AI.