marcelfahle
Joined 16 years ago
Joined 16 years ago
Video intelligence platform for coaching programs and training companies. The problem: these businesses sit on 200-500+ hours of video content that becomes a "content graveyard" - students can't find what they need, coaches burn out answering the same questions, churn stays high.
We do deep transcript + metadata extraction, then layer RAG search and an AI assistant that can answer questions with timestamped citations back to the exact video moment. Think "ChatGPT for your video library" but with accurate sources instead of hallucinations. Tech: Phoenix/Elixir backend, Next.js portals, two-tier RAG architecture.
Currently serving a few coaching programs in high-touch sales mode. Would love feedback from anyone who's built RAG systems over media content - curious how others handle the signal extraction problem (transcripts are noisy, you need to identify what's actually being taught vs filler).
Excuse the question, but "forced alignment" is when you don't have timestamps, like in webvtt?
Speaker 0 00:00:12 Oh, fuck motherfucker. Okay. Fuck, fuck, fuck, fuck, fuck, fuck, fuck, fuck.
My little fuck.
Speaker 1 00:02:10 Oh, fuck. Oh, fuck,
Speaker 0 00:02:25 Fuck, fuck, fuck, fuck, fuck, fuck, fuck, fuck my motherfucker.
Speaker 1 00:02:53 Fucking a.
Speaker 0 00:02:54 Mm-hmm. <affirmative> motherfucker. Fuck me. Um,> Whether the vehicle was driven by the 43-year-old driver or not is the subject of the investigation and is currently unknown, a police spokesperson said on Tuesday. The cause of the accident remains unclear.