Location Station careers

Location Station is an “AI radio” for your city: a daily spoken briefing you can talk to. It pulls from local & global newsrooms, newsletters, event feeds, and public data, then writes, voices, and streams a short program with citations. Every minute listened—and each conversational follow-up that quotes a source—is attributed to the original publishers for payout.

Hard problems we’re solving (and how):

Attribution & payouts: map each sentence and quoted snippet to sources with timestamps and confidence to generate an auditable ledger based on attention minutes.

Truth & provenance: maintain a citation chain from ingestion → summarization → script → audio; the agent can expose or reopen sources on demand.

Latency & cost: just-in-time generation, caching, incremental updates, ffmpeg multitrack mix; HLS/WebRTC delivery; evals on quality, WER, latency, $/min.

Personalization that respects place: tagging + retrieval tuned to geography and interests; learn from completions/skips and thumbs.

Local data sparsity/messiness: unify sites, newsletters, calendars/APIs; dedupe events, disambiguate entities, normalize metadata for ranking.

Mobile & scale: offline caching, background audio, “continue/go deeper” prompts, fast resume; templated ingestion + publisher onboarding to light up new metros efficiently.

Status
Prototype pieces running: ingestion + GPT-tagging, retrieval, script composer, TTS, multitrack mix, 10-minute demo slice. Early publisher outreach is underway.

Stack
Node/TypeScript; AWS (EC2, WAF, ELB, VPC); DynamoDB; OpenAI Assistants; ElevenLabs TTS; ffmpeg multitrack; HLS/WebRTC; clients in Kotlin/Android, iOS, and React.

Why now
Local news needs a new economic model, and AI voice + retrieval are finally good and cheap enough to deliver a trustworthy, habit-forming newscast that pays journalists for attention—including conversational minutes.