Products / 10

Media · AI content pipeline · Beta

MTB Feed

A news feed engine for mountain biking.

MTB Feed downhill news feed

A multi-tenant news feed engine. It pulls from Instagram, RSS and YouTube, drafts stories with an LLM, runs polls, and puts everything through an admin moderation queue before it goes live.

What it does

  • Instagram, RSS and YouTube fetchers
  • LLM-drafted stories with human moderation
  • Polls and per-discipline feeds
  • Multi-tenant: one engine, many feeds

What we built

MTB Feed is a news engine we designed and built end to end: one deployment runs any number of topic feeds, each with its own sources, branding and rules. The first feed covers gravity mountain biking: downhill and enduro riders, teams and brands.

Every few hours it checks Instagram, RSS, YouTube, team sites and press pages, removes duplicates, scores each item for relevance and importance (wins, podiums, signings, injuries, launches), and groups reports of the same event. An LLM then writes a short story with links back to the sources, and editors approve, pin or edit it in the admin before it goes live.

How it is built

A React front end with Cloud Functions (Node) and Firestore behind it. Fetchers are pluggable, each source tracks its own health with automatic backoff, and the scheduler runs feeds within a shared time budget, with a lock so two runs never overlap.

The LLM layer works with several providers and falls back to an extractive writer, so a failed API call never empties the feed. We read Instagram only through compliant routes (the official Graph API and oEmbed), always link to the original post and do not rehost creators’ media.

Want this level of product and engineering on yours?

Book a call

MVP to Production