Back to Our Work

Real-time media intelligence

StreamPulse

A unified streaming command center that aggregates localized entertainment news, tracks global media releases, and curates personalized user watchlists across multiple international regions.

Built as a Vite and React single-page application backed by Vercel serverless functions and a Neon PostgreSQL database via Drizzle ORM. The core engine runs an automated multi-region RSS ingestion pipeline triggered by Vercel Crons. To control operational overhead and gate LLM token spend, incoming articles undergo strict pre-spend title and URL deduplication before being dynamically classified by an OpenRouter structured AI engine, safely executing within tight serverless function time budgets.

  • Real-time Data
  • Content Aggregation
  • Gemini API
  • AI-Accelerated Delivery

Please provide your project scope and details in the form so our team can best prepare for the session.

https://streampulse.semsotai.ca/dashboard
StreamPulse dashboard showing browse releases, filters, and media grid

System Architecture & Engineering Highlights

Autonomous AI Ingestion

Daily Vercel Crons sweep multi-region RSS feeds across the US, HK, TW, and IN. Incoming articles are duplicate-hashed before spend, then summarized and categorized by an OpenRouter LLM pipeline within serverless time budgets.

Hardened Edge Resiliency

React 18 Strict Mode–safe auth token persistence via sessionStorage keeps sessions stable across remounts. Cross-origin image fallback boundaries keep poster grids intact on mobile Safari and Chrome when remote CDNs fail.

Cloud-Native Stack

Vite + React SPA with Tailwind CSS, deployed on Vercel Edge Serverless Functions and backed by a Neon Serverless PostgreSQL schema via Drizzle ORM for low-latency, region-aware data access.