#business #idea # [[Epistemic status]] #shower-thought #to-digest # Related - [[Ideation]] - [[Start-up ideation]] # TODO > [!TODO] TODO # Start-up idea Supabase for edge machine learning models ## Problem/opportunity - through last decades we go from big, server-deployed, slow [[Artificial intelligence|AI]] models to edge, fast, UX integrated models (OCR, TTS, STT, etc) - there are still no services that handle the end-to-end process of: - training/fine-tuning/context-specializing models - deploying on edge to specific users, specific models, 0 pain, 1 line of code ### Concerns - ML is going from FT to rather [[Prompt]] engineering, #prompt hacks - less need to FT? ## Solution ## Markets/customers/users market: https://firebase.google.com/docs/ml ## Why - Popular: increasingly using [[Software 2.0]] and [[Software 3.0]] on edge - Growing: increasingly using [[Software 2.0]] and [[Software 3.0]] on edge - Urgent: no end-to-end solution - hacks - download - ft - deploy etc - data privacy hacks - Expensive: lot of infra to manage - Mandatory: - [[Founder-market-fit|Why do you want to solve this problem?]] experienced pain - cool/fun ? - [[What's your unfair advantage?]]: little xp with edge ml - lot of xp with mobile dev -