Sign up to save your podcasts
Or
Share UK AI Startups Leaving for SF! Inside London's Coolest AI Lab!
Share to email
Share to Facebook
Share to X
Share to email
Share to Facebook
Share to X
Or
UK AI Startups Leaving for SF! Inside London's Coolest AI Lab!
Meet the Team behind Valyu AI: Harvey & Hirsh!website: https://valyu.aiplatform (to get apikey and test the product): https://platform.valyu.network (will be platform.valyu.ai)Twitter page: https://x.com/ValyuNetwork📲 Get premium wireless for as low as $15 a month with my partner at https://www.mintmobile.com/singhMy Recommendations: https://linktree.com/singhinusa1. How I Edit podcast & enhance audio: https://adobe.prf.hn/click/camref:1100l568ad/destination:https%3A%2F%2Fpodcast.adobe.com2. Checkout KodeKloud & Get upto 45% Off https://bit.ly/4dpsE7y (includes AWS, GCP, Azure Subscriptions with the Pro Plan)
00:00 — Intro: Value AI’s mission beyond the web. 01:02 — Where the team is from. 01:33 — London vs SF: ecosystem energy. 02:45 — “What part of the body is London?” Appendix joke. 03:09 — Wild founder stories in London. 03:52 — Pull-up bar challenge. 04:37 — Founders living together: routines and work culture. 05:07 — Work cadence: “8 days a week.” 05:33 — Dating life: social commentary. 06:21 — UK brain drain and travel. 6:39 — Co-founders’ deep technical backgrounds (start of deep dive). 07:03 — Hrik’s journey: Kenya to UCL, compilers. 08:46 — PhD in federated learning explained. 09:44 — AlphaGo/UCL link: learning RL from David Silver. 10:34 — Harvey’s path: agentic systems and retrieval bottlenecks. 11:39 — Third co-founder Alex: scaling vector search. 12:25 — First-principles approach to IR. 13:00 — Unique edge: non‑web and hard-to-access data. 13:38 — Reinventing IR with LLM-era tooling. 14:27 — Curation quality: papers, patents, expert sources. 14:59 — Product focus: content over links/snippets; partnerships. 15:38 — Finance vertical: real-time market data, SEC filings. 16:24 — Legal and science: patents, case law, pharma research. 17:24 — “Mars” goal: become AI’s information layer for knowledge work. 18:27 — Scaling peta‑byte IR: low-level systems, latency, cost. 20:07 — Building own vector DB and retrieval stack; compounding optimizations. 21:23 — Heterogeneous indices + orchestration layer. 22:43 — “Deepseek moment of search” positioning. 23:20 — Motivation: curiosity and building. 24:03 — Meeting internet pioneer Peter Kirstein at UCL. 26:10 — Haskell sprint to build a compiler; confidence loop. 28:12 — Expertise summary: distributed systems vs agentic search. 29:10 — Startup timeline: late 2023 research to product. 29:56 — CSX story: chance meeting → accelerator. 32:13 — Pivot: training data → inference/search layer. 33:33 — Why AI is “easy” but search is hard: cost and optimizations. 34:24 — Launch tease: benchmarks coming next week. 35:04 — Designing for agents: tool-call mode, agent UX. 37:02 — Misconception: Web search built for humans/SEO, not AI. 38:18 — Why London: research talent, UCL crucible. 40:04 — Advice: explore accelerators; visit SF for perspective. 42:06 — Next steps: moving to SF soon; closing thanks. My Gear:-📕 Favorite Book for Software Engineers: https://geni.us/GT9jvz💻 My Gear Page: https://singhinusa.com/my-tech/📱iPhone: https://geni.us/J9c6GoY📸Sony Camera: https://geni.us/zShMGT4⿍ Tripod: https://geni.us/iVsd🎙Vlogging & Podcast Mic https://geni.us/v915Z9H~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Follow me on - ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Follow me on - Instagram - https://www.instagram.com/iHarnoorTwitter - https://x.com/iHarnoorSinghLinkedin: https://www.linkedin.com/iHarnoorE-mail for BUSINESS INQUIRY & HELP- [email protected]
View all episodes
By Harnoor SinghMeet the Team behind Valyu AI: Harvey & Hirsh!website: https://valyu.aiplatform (to get apikey and test the product): https://platform.valyu.network (will be platform.valyu.ai)Twitter page: https://x.com/ValyuNetwork📲 Get premium wireless for as low as $15 a month with my partner at https://www.mintmobile.com/singhMy Recommendations: https://linktree.com/singhinusa1. How I Edit podcast & enhance audio: https://adobe.prf.hn/click/camref:1100l568ad/destination:https%3A%2F%2Fpodcast.adobe.com2. Checkout KodeKloud & Get upto 45% Off https://bit.ly/4dpsE7y (includes AWS, GCP, Azure Subscriptions with the Pro Plan)
00:00 — Intro: Value AI’s mission beyond the web. 01:02 — Where the team is from. 01:33 — London vs SF: ecosystem energy. 02:45 — “What part of the body is London?” Appendix joke. 03:09 — Wild founder stories in London. 03:52 — Pull-up bar challenge. 04:37 — Founders living together: routines and work culture. 05:07 — Work cadence: “8 days a week.” 05:33 — Dating life: social commentary. 06:21 — UK brain drain and travel. 6:39 — Co-founders’ deep technical backgrounds (start of deep dive). 07:03 — Hrik’s journey: Kenya to UCL, compilers. 08:46 — PhD in federated learning explained. 09:44 — AlphaGo/UCL link: learning RL from David Silver. 10:34 — Harvey’s path: agentic systems and retrieval bottlenecks. 11:39 — Third co-founder Alex: scaling vector search. 12:25 — First-principles approach to IR. 13:00 — Unique edge: non‑web and hard-to-access data. 13:38 — Reinventing IR with LLM-era tooling. 14:27 — Curation quality: papers, patents, expert sources. 14:59 — Product focus: content over links/snippets; partnerships. 15:38 — Finance vertical: real-time market data, SEC filings. 16:24 — Legal and science: patents, case law, pharma research. 17:24 — “Mars” goal: become AI’s information layer for knowledge work. 18:27 — Scaling peta‑byte IR: low-level systems, latency, cost. 20:07 — Building own vector DB and retrieval stack; compounding optimizations. 21:23 — Heterogeneous indices + orchestration layer. 22:43 — “Deepseek moment of search” positioning. 23:20 — Motivation: curiosity and building. 24:03 — Meeting internet pioneer Peter Kirstein at UCL. 26:10 — Haskell sprint to build a compiler; confidence loop. 28:12 — Expertise summary: distributed systems vs agentic search. 29:10 — Startup timeline: late 2023 research to product. 29:56 — CSX story: chance meeting → accelerator. 32:13 — Pivot: training data → inference/search layer. 33:33 — Why AI is “easy” but search is hard: cost and optimizations. 34:24 — Launch tease: benchmarks coming next week. 35:04 — Designing for agents: tool-call mode, agent UX. 37:02 — Misconception: Web search built for humans/SEO, not AI. 38:18 — Why London: research talent, UCL crucible. 40:04 — Advice: explore accelerators; visit SF for perspective. 42:06 — Next steps: moving to SF soon; closing thanks. My Gear:-📕 Favorite Book for Software Engineers: https://geni.us/GT9jvz💻 My Gear Page: https://singhinusa.com/my-tech/📱iPhone: https://geni.us/J9c6GoY📸Sony Camera: https://geni.us/zShMGT4⿍ Tripod: https://geni.us/iVsd🎙Vlogging & Podcast Mic https://geni.us/v915Z9H~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Follow me on - ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~Follow me on - Instagram - https://www.instagram.com/iHarnoorTwitter - https://x.com/iHarnoorSinghLinkedin: https://www.linkedin.com/iHarnoorE-mail for BUSINESS INQUIRY & HELP- [email protected]