Narad - GenAI Powered News Intelligence Platform

Turning global news into clear, actionable intelligence with Generative AI.

Comment

Inspiration

We were inspired by the growing problem of information overload, misinformation, and conflicting news narratives. We wanted to build more than a conventional news aggregator: an intelligent platform that connects stories, verifies claims, and helps people understand the bigger picture behind the news. Inspired by Narad Muni, the messenger of information and wisdom, we named our platform Narad.

What it does

Narad is a GenAI-powered multilingual news intelligence platform covering 80+ news sources across 10+ Indian languages. Its key capabilities include:

  • AI-powered claim verification and evidence scoring.
  • Narrative conflict detection and cross-source validation.
  • Multilingual news aggregation and semantic search.
  • Event intelligence and causal relationship exploration.
  • Sentiment and source bias analysis.
  • AI-generated briefings, fact sheets, and conversational news exploration through Ask Narad.
  • India Command Center with geographical and state-level insights.

Narad transforms scattered news articles into interconnected, actionable intelligence.

How we built it

We developed Narad using a Next.js and React frontend with a FastAPI backend. Our automated ingestion pipeline collects and processes news through RSS feeds and APIs, while spaCy and sentence-transformers handle NLP and semantic extraction.

We integrated FAISS and PostgreSQL with pgvector for vector search and retrieval. Amazon Bedrock provides access to multiple foundation models, including Llama 3.3 70B, DeepSeek, and Amazon Nova Pro. We also implemented Retrieval-Augmented Generation (RAG) to generate contextual, evidence-grounded responses. Deployment uses AWS Amplify, S3, and CloudWatch.

Challenges we ran into

Our biggest challenges included processing multiple Indian languages, removing duplicate articles, and retrieving relevant information from large news datasets. Building reliable claim verification and detecting conflicting narratives required advanced semantic comparison rather than simple keyword matching. We also faced challenges in coordinating multiple LLMs, maintaining response consistency, and optimizing system performance.

Accomplishments that we're proud of

We successfully built and deployed a multilingual news intelligence platform with multiple interconnected AI modules. We are proud of implementing semantic retrieval, evidence-based claim analysis, narrative conflict detection, and geographical intelligence capabilities. Our biggest achievement was securing All India 2nd Position at the AI for Bharat Hackathon 2026.

What we learned

We gained practical experience in multilingual NLP, RAG pipelines, vector databases, semantic search, event intelligence, and LLM orchestration. We learned that building reliable AI systems requires more than integrating an LLM; it requires strong data pipelines, relevant evidence retrieval, and carefully designed intelligence workflows.

What's next for Narad

We aim to improve multilingual intelligence, strengthen evidence-based verification, and develop temporal event tracking and predictive early-warning capabilities. We also plan to expand our news coverage globally and build advanced knowledge graphs connecting events, people, organizations, and claims.

Our long-term vision is to transform Narad into a comprehensive global intelligence platform that helps people understand not just what is happening, but why it matters.

Built With

Share this project:

Updates

Submission history