NewsInsight.ai

NewsInsight.ai is an autonomous news analyst: it filters spam, scores credibility, extracts entities, and delivers verified, sentiment-tagged summaries in seconds—turning noise into knowledge.

Comment

Inspiration

I was driven by the overwhelming information problem—over 2.5 million news articles are published daily, often mixed with ads, clickbait, and misinformation. As a developer, I personally experienced the frustration of information overload, inconsistent quality, lack of emotional context, and the impracticality of manual curation. I recognized the potential of AWS Bedrock's Claude 3.5 Haiku to create an autonomous AI agent capable of intelligently curating and analyzing news while providing rich contextual insights.

What It Does

NewsInsight.ai is a fully autonomous AI agent that revolutionizes news consumption through:

  • Intelligent curation across 15 trending categories with multi-layer AI filtering (50% acceptance rate)
  • Advanced AI analysis, including sentiment analysis, entity recognition, and quality scoring, utilizing AWS Bedrock
  • Real-time search with sub-second response times and an 85% cache hit rate
  • 24/7 autonomous operation with self-optimization and automatic cleanup
  • Interactive AI explanations, allowing users to ask the AI to clarify any article

How I Built It

I developed a sophisticated multi-layer AI agent using:

  • AI Core: AWS Bedrock Claude 3.5 Haiku for four-layer decision making
  • Frontend: React 18 and Tailwind CSS deployed on Vercel with a newspaper-inspired design
  • Backend: FastAPI on Railway with smart caching and content filtering
  • Data: DynamoDB and S3 for optimized storage and queries
  • Automation: AWS Glue, Lambda, and EventBridge for nightly ETL processing
  • APIs: NewsAPI and Guardian API for global news aggregation

Challenges I Encountered

  • Complexity of AWS Bedrock required specific prompt engineering for consistent JSON responses
  • Multi-layer filtering posed challenges in balancing quality and speed while avoiding false positives
  • Achieving real-time performance meant I had to ensure sub-second responses during AI analysis
  • Deployment issues with Railway included health check complications related to AWS credentials configuration
  • EventBridge integration necessitated a Lambda bridge for Glue job triggers
  • Ensured the newspaper-inspired design worked seamlessly across all devices

Accomplishments I'm Proud Of

  • Successfully built a truly autonomous AI agent requiring no human intervention
  • Exceptional performance with sub-second search times, an 85% cache hit rate, and a 50% cost reduction
  • Professional user experience with a newspaper-inspired design and seamless dark mode
  • Production-scale architecture with live deployment serving real users
  • Measurable impact, curating 50-200 articles daily with a 50% acceptance rate
  • Mastery of complex AWS integrations with six services working together seamlessly

What I Learned

  • AI Development: Prompt engineering is crucial, and multi-layer reasoning is more effective
  • AWS Architecture: While Bedrock is powerful, it requires careful handling; EventBridge and Lambda can create flexible systems
  • Production Deployment: Managing environment variables is tricky, and monitoring is essential
  • Frontend Development: Using React and Tailwind is productive, but implementing dark mode requires upfront planning
  • Data Processing: Content filtering is more challenging than expected; hybrid real-time/batch approaches tend to work best

What's Next for NewsInsight.ai

  • Short-term (3 months): Implement multi-language support, trend prediction, bias detection, user personalization with AI recommendations, and achieve sub-100ms response times.
  • Medium-term (3-12 months): Expand to a global multi-region deployment, create mobile apps, develop a browser extension, launch an API marketplace, and evolve into a multi-agent system.

Built With

Share this project:

Updates