NewsLens - AI Powered Personalized News Aggregator

A smart news platform that aggregates articles from 100+ sources and uses AI to deliver a personalized, bias-aware feed. Instead of keeping users inside an echo chamber, NewsLens surfaces the same story from multiple outlets side by side so readers always get the full picture. Think of it as a self-serve Flipboard meets ground News with an AI layer on top that actually understands what you care about.

The Problem It Solves

Most news apps trap users in filter bubbles by showing only content that matches their existing views
Reading the same story from one outlet gives an incomplete and often skewed perspective
There is no single platform that combines aggregation, AI summarization, and bias labeling together
Managing multiple news subscriptions across different apps is fragmented and time-consuming
Readers have no easy way to know how much news they are actually consuming or whether sources lean left or right
College students and researchers need multi-perspective reading without paying for five different subscriptions

Tech Stack

Frontend: React.js, TailwindCSS
Backend: Node.js with Express.js for the API layer
Scraping and NLP Pipeline: Python with BeautifulSoup and NewsAPI for content ingestion and processing
Primary Database: PostgreSQL for articles, users, preferences, and reading history
Search: Elasticsearch for full-text article search with fast indexing
Cache: Redis for trending articles, feed caching, and session management
AI: OpenAI API for article summarization, bias detection, and topic clustering
Auth: Google OAuth with JWT for session handling
Email: SendGrid for weekly digests and notification emails
Deployment: AWS EC2 for backend, Docker for containerized services, cron jobs for the scraping pipeline

Core Features (MVP)

Personalized Feed
- AI model learns from reading history, liked topics, bookmarked articles, and explicitly skipped content
- Feed refreshes every 30 minutes with newly scraped articles ranked by personal relevance score
- Onboarding flow asks users to pick 5 topics and 3 sources to cold-start the recommendation engine
- Toggle between Personalized, Trending, and Latest modes without losing preferences
- Thumbs up and thumbs down signals on articles to tune the feed in real time
Multi-Source Story View
- Clustering algorithm groups articles covering the same event from different outlets
- Same story shown from 3 to 5 different publications in a side-by-side card layout
- Highlights how different outlets frame the same facts through headline and tone comparison
- Source trust score displayed alongside each outlet based on media reliability databases
- Users can expand any card to read the full article on the original publication
AI Summaries
- One-paragraph TL;DR generated by OpenAI GPT-4o for every article in the feed
- Summary highlights the who, what, when, where, and why in under 100 words
- Original article always accessible below the summary for users who want full context
- Summary generation runs asynchronously in the scraping pipeline before articles reach the feed
- Users can flag a summary as inaccurate and it gets re-queued for regeneration
Bias Indicator
- Every source tagged with a political lean: Left, Center-Left, Center, Center-Right, or Right
- Bias labels sourced from AllSides Media Bias Ratings and Ad Fontes Media
- Color-coded label appears on every article card so the lean is visible at a glance
- Organizer dashboard view shows the political distribution of a user's reading history
- Bias filter lets users intentionally choose to read only centrist sources or explore all leans
Topic Bundles
- Users follow curated topic streams like India Elections, AI News, Climate, or Cricket
- Each bundle is a named feed powered by keyword clusters and NLP topic classification
- New bundles auto-suggested based on trending cluster activity in the last 48 hours
- Bundle page shows top stories, most-read articles, and active sources in that topic
- Users can create a custom bundle by combining keywords and adding specific sources

Advanced Features (Scalable Ideas)

Fact-Check Layer
- Disputed claims in articles are flagged automatically using integration with PolitiFact and Snopes
- Flagged sentences shown with a yellow highlight and a link to the relevant fact-check
- Fact-check coverage score per outlet shows how often their claims are disputed or verified
- Users can manually submit a claim for fact-checking from any article in the feed
Reading Time Tracker
- Tracks total minutes spent reading per day with a daily breakdown chart
- Shows how many articles read, sources visited, and topics consumed in a session
- Weekly reading report sent every Sunday with a breakdown of time and topic distribution
- Daily reading goal feature with a progress ring on the homepage
Weekly Digest
- Auto-generated Sunday morning email summarizing the 10 biggest stories from the past week
- Digest is personalized based on the user's topic preferences and reading history
- Each story in the digest includes a two-sentence AI summary and a link to the full multi-source view
- Users can customize digest delivery time and choose between daily or weekly cadence
Offline Reading
- Save any article to a personal reading list accessible without an internet connection
- Saved articles downloaded as clean reader-mode HTML with images stripped for speed
- Sync reading progress across devices so half-read articles resume from the right position
- Folder organization for saved articles with custom labels like Research, Share Later, or Long Read
Journalist Profiles
- Follow individual reporters across publications regardless of which outlet they currently write for
- Journalist profile page shows bio, beat, recent articles, and bias profile of their coverage
- Alerts when a followed journalist publishes a new piece even if it appears in an unfamiliar outlet
- Readers can see which journalists consistently cover a specific topic bundle
News Quiz
- Weekly quiz of 10 questions automatically generated from the articles a user read that week
- Multiple choice format with answers pulled directly from article content
- Score tracked over time with a leaderboard among friends or followers
- Share quiz results on social media with a summary of the week's top stories

Monetization Strategy

Free plan — 20 articles per day, standard feed, AI summaries, bias labels, no offline reading
Reader plan — unlimited articles, offline reading, weekly digest, reading tracker, ad-free experience
Pro plan — journalist profiles, custom topic bundles, fact-check layer, export reading history, priority support
News organization API — white-label feed and summarization API sold to media companies and publishers
Featured topic bundles — brands or event organizers pay to sponsor a curated bundle during major events

Why This Project Stands Out

Combining a scraping pipeline, NLP classification, OpenAI summarization, and bias detection in one system is genuinely hard and rare in student portfolios
Multi-source story clustering using cosine similarity or topic modeling shows real ML thinking beyond basic CRUD
Redis caching strategy for a high-volume feed is an interview-ready system design talking point
The product is immediately usable as a real consumer app targeting news-literate urban readers in India
End-to-end thinking from content ingestion through personalization to post-reading surveys makes this a portfolio-defining project that shows full product sense

NewsLens - AI Powered Personalized News Aggregator

Domains

Technologies