Building the World's Most-used Medical AI | Dr. Thomas Kelly of Heidi - E716

Dr. Thomas Kelly was a vascular surgery trainee in Melbourne before he walked away from medicine to fix the problem he hated most: paperwork stealing time from patients. Today he is co-founder and CEO of Heidi Health, an AI medical scribe and "care partner" used by clinicians in more than 100 countries, valued at US$465M after its 2025 Series B led by Point72. In this episode, Jeremy Au and Dr. Tom break down how GPT-4 nearly destroyed the company, forcing a brutal pivot from an AI model company into a product company, and how that near-death moment led to becoming one of the world's most-used healthcare AI tools. They dig into why great AI scribing is far harder than a demo, how Heidi is expanding into clinical decision support and medical-device territory, why he isn't scared of OpenAI in healthcare, and why Singapore, Japan, and the rest of Asia are core to the roadmap. A must-watch for founders, venture capitalists, and operators building AI, health tech, and vertical software across Southeast Asia. Watch, listen or read the full insight at https://www.bravesea.com/blog/heidi-health-tom-kelly  BRAVE is Southeast Asia's leading tech podcast, hosted by Jeremy Au. Honest conversations with the region's top founders, investors, and operators on building startups in Southeast Asia. New episodes every week. Subscribe so you never miss one. Listen & Subscribe YouTube (English), YouTube (Bahasa Indonesia), Spotify (English), Spotify (Bahasa Indonesia), Spotify (Chinese), Spotify (Vietnamese), Apple Podcasts Follow BRAVE LinkedIn, X (Twitter), Instagram, TikTok, WhatsApp Follow Jeremy Au LinkedIn, X / Twitter, Instagram, TikTok, Facebook, Threads, Twitch Resources Get transcripts, startup resources & community discussions at www.bravesea.com #HealthcareAI #AIScribe #FounderStory #AIinHealthcare #StartupPivot #SoutheastAsia #VentureCapital 00:00 – From vascular surgery to AI: meet Dr. Tom Kelly 01:38 – Why he became a doctor: the "platonic ideal" GP 02:55 – Maths, machine learning, and the pre-med detour 04:22 – GAMSAT YouTube videos to his first business 07:06 – Building Oscer, an early AI clinical tutor 08:58 – Why doctor paperwork is a capacity crisis 13:05 – Is an AI scribe easy or hard to build? 17:35 – From scribe to full AI care partner 18:28 – Clinical AI and medical device regulation 21:32 – Roadmap: tasks, primitives, and on-prem hardware 26:24 – Going global: Singapore, Japan, and Asia 28:50 – Localizing for Malay, Hokkien, and Singlish 32:52 – Why clinicians drive the innovation 34:21 – The GPT-4 pivot that nearly ended Heidi 37:20 – Competing with OpenAI and how to run a pivot