BritChips podcast: RayD Technologies - Learning with Limited Data

BritChips Podcast: Cosine - A new angle on Sovereign AI

“What’s the dream?” I asked Yang Li, cofounder and COO of AI code generator startup Cosine. “We want to be ten-times bigger than Deep Mind,” came the response. Of course you do, I thought to myself - and you might too. But can I encourage you to watch/listen to the full 30-ish minute episode of The BritChips Podcast: Cosine - A New Angle on Sovereign AI before making up your mind. You see, the market that Cosine is addressing is not all the new stuff. It’s all the old stuff, the legacy code that has been deeply embedded in commercial enterprises and public sector bodies for years if not decades. And often there’s nobody left in the organisation who really understands how it works, making maintenance and upgrades an arduous and highly risk-prone exercise. Yang believes they have the answer - an AI codegen specifically designed to maintain and enhance applications coded in arcane - even archaic - languages. And Cosine has a USP that Yang believes the ‘usual suspect’ AI codegens can’t beat. They see Cosine as the only end-to-end sovereign AI coding platform built, owned, and operated entirely in Britain. The UK government is clearly a believer. Cosine was one of the first startups to be accepted into the government’s £500m Sovereign AI initiative, awarding it access to Isamabard, the UK’s most powerful supercomputer. So, see what you think. Podcast summary Discover how UK-based startup Cosine is leveraging AI to transform legacy code management across defence, finance, and semiconductor industries. Cosine cofounder and COO Yang Li explains their innovative approach, funding journey, and ambitious future plans to make Britain a leader in sovereign AI. * Yang Li introduces Cosine and their mission to build the first end-to-end sovereign AI coding platform in Britain * The origins of Cosine and how a chance meeting at a wine bar sparked the project * Differentiating between greenfield and brownfield projects in AI coding applications * The technical approach behind Cosine’s language models trained specifically on legacy languages like COBOL and VHDL * How Cosine’s platform enhances maintenance and feature addition in critical sectors * The UK government’s £500 million sovereign AI initiative and its impact on Cosine’s growth * Current customer deployment status and future industry targets, including semiconductor manufacturing * Funding journey: bootstrapping, YC acceleration, US and UK investment, and strategic partnerships * The long-term vision: making UK AI as influential as DeepMind and aiming for global leadership * The role of Form1 Partners in supporting Cosine’s go-to-market strategy Topics * Yang Li’s background and Cosine’s origin story * Differentiating greenfield vs brownfield projects * Cosine’s approach to legacy system maintenance * Use cases in defense, finance, and critical infrastructure * Ensuring AI decisions are audit-trailed for regulation * Entering the semiconductor language market * Training models for low-level hardware languages * Collaborations with chip industry players * Funding origins: bootstrap, Y Combinator, and investments * UK government’s sovereign AI program * The capabilities enabled by government support * Current team size and growth plans * Key milestones: training a sovereign large language model * Future vision: positioning Cosine as a UK AI leader * How Cosine engaged with Form1 Partners The BritChips Podcast is proudly sponsored by Silicon Catalyst UK, Official Partner of the government-funded ChipStartUK semiconductor incubator. We gratefully acknowledge the support of Form1 Partners in the making of this podcast. Form1, for which Anthony Miller is an advisor, is the fast-growing commercialisation partner for B2B tech scale-ups and established corporates. See here [https://form1partners.com/] for more information about Form1. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit anttheantidote.substack.com [https://anttheantidote.substack.com?utm_medium=podcast&utm_campaign=CTA_1]

BritChips Talking Heads: Mark Isaacs, HarwellXPS

Welcome to the first episode of Talking Heads, a new BritChips series of short-form videos featuring people who are helping to support UK semiconductor startups. My first ‘Talking Head’ is Mark Isaacs, who leads HarwellXPS, the Engineering and Physical Sciences Research Council’s national facility for photoelectron spectroscopy, based at the government’s Harwell research campus. This world-class facility is engaged in surface analysis of advanced materials such as semiconductors in order to detect microscopic defects which could cause the device to underperform or fail completely. After listening to this clip you might want to swing over to The BritChips Podcast: Deep Fabrication - Predicting Defects in which I am in conversation with Dr. Yasir Noori, cofounder and CEO of London-based semiconductor design tool startup, Deep Fabrication, which is developing AI tools that can predict defects in semiconductor designs prior to manufacturing. After all, they do say that prevention is better than cure! Talking Heads is also available on the BritChips YouTube channel (https://www.youtube.com/@BritChips) and in audio-only form on Apple Podcasts and Spotify. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit anttheantidote.substack.com [https://anttheantidote.substack.com?utm_medium=podcast&utm_campaign=CTA_1]

BritChips Podcast: Chipletti - Let's Get Physical!

I first spoke to Chipletti cofounders Kauser Johar and Catriona Wright in The BritChips Podcast ‘Chipletti’s Big Stack’ in June last year. Chipletti, you may recall, was one of the startups selected to participate in the third cohort of the government-sponsored ChipStart incubator, managed by Silicon Catalyst UK. At that time they were hard at work developing high-speed cache memory chiplets (the clue is in the company name) designed to stack on top of logic chips so you get more cache memory without needing extra space on the silicon. But that’s all changed as you’ll hear (and see) in this latest episode of The BritChips Podcast, in which they explore the latest developments in edge AI hardware, focusing on innovative memory stacking, chiplet integration, and simulation tools that are revolutionizing the design and deployment of AI inference systems at the edge, aka Physical AI. In this episode: * The progression from cache memory optimization to edge AI applications * Challenges and solutions in 3D memory stacking for low-power devices * Designing chiplets for AI inference in mobile and real-time applications * The role of simulators in predicting hardware performance before silicon fabrication * Milestones: FPGA prototyping, chiplet integration, and silicon manufacturing timelines * The strategic focus on supply chain sovereignty and local manufacturing for AI hardware * How this integrated approach uniquely positions the company within a competitive market Topics: * Introduction to the evolution from cache memory to edge AI inference hardware * Detailed benchmarking results on memory stacking performance improvements * Bridging cache memory work with new edge AI chip architectures * Exploring applications suitable for 10-15 watt power envelopes, including drones and robots * Real-time AI inference and the importance of latency in safety-critical systems * GPU limitations in real-time environments and architecture optimization for edge AI * Designing chips that are more resilient, reliable, and power-efficient for edge AI use * Transition from memory chiplet design to integrated compute chiplets * The move from selling memory chiplets to providing complete system modules * Development of a high-fidelity simulator for hardware performance prediction * Creating the simulator with internal expertise and academic collaboration * Engaging potential customers and enabling early deployment through simulation * Making the simulator accessible to developers early in the hardware development cycle * How the simulator models future technology and non-existent hardware * The abstraction level and accuracy of the simulation models * Using simulation to inform design decisions across multiple future hardware scenarios * Customization and flexibility in the simulator for different hardware architectures * Commercial use: Charged for simulation or a strategic tool? * Milestones: FPGA prototypes, module integration, and upcoming silicon tape-outs * Industry-standard FPGA development plans and timelines for demonstrations * Funding strategies and milestone timelines, including venture capital and grants The BritChips Podcast is proudly sponsored by Silicon Catalyst UK, Official Partner of the government-funded ChipStartUK semiconductor incubator. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit anttheantidote.substack.com [https://anttheantidote.substack.com?utm_medium=podcast&utm_campaign=CTA_1]

BritChips podcast: RayD Technologies - Learning with Limited Data

In this latest episode of The BritChips Podcast, Rayd Technologies cofounder and CEO Kilian Stenning explains how he, co-CTOs Riccardo Sapienza (Professor of Physics at Imperial) and Kirsten Moselund (Head of Laboratory at PSI Paul Scherrer Institut) and CSO Jack Gartside (Associate Professor of Physics at Imperial) are leading pioneering work on photonic semiconductor technology and novel algorithms with the potential to shape the future of AI. In this episode: * Killian’s unexpected journey from nuclear accident analysis to semiconductor startup founder * The origins and evolution of RayD Technologies and its unique name * How brain-inspired dynamics enable learning from limited data * The role of photonics and nonlinear systems in ultra-fast AI processing * The current hardware prototype and plans for miniaturization * Application focus: medical imaging, robotics, and beyond * Funding progress, including support from the Royal Academy of Engineering * The startup’s strategic collaborations and upcoming milestones * Insights on balancing technological deep dives with business growth * The importance of team, network, and perseverance in startup success Timestamps (approx) 00:00 - Introduction to Kilian and the fusion of photonics and brain-inspired computing 02:05 - Kilian’s early career in nuclear accident analysis in Japan 03:04 - How photonics hardware inspired the company’s research direction 04:11 - Developing scalable semiconductor photonic platforms at Imperial College 05:24 - The collaboration between hardware and algorithm teams to create RAYD’s technology 06:45 - From academic research to commercial startup: the journey of recognition 07:01 - Key breakthroughs in processing medical images with limited data 08:23 - The story behind the company’s name, RAYD 09:47 - Differentiating RAYD’s neuromorphic approach from traditional AI and neuromorphic systems 10:34 - Brain-inspired features like parallelism and nonlinear activation functions 11:41 - How the system learns efficiently, mimicking human learning with minimal data 12:46 - Achieved reductions of up to 30x in training data requirements 13:26 - Hardware versus algorithms: the photonics platform’s critical role 14:22 - Unique nonlinear photonic processing in RAYD’s device, using laser systems 15:31 - Application scenarios in medical imaging and robotics, sourcing data from public datasets 17:22 - Development stage: tiny device, large supporting hardware, and plans for miniaturization 18:27 - Roadmap to portable systems and real-world testing 19:14 - Interest from robotics companies for on-device learning and adaptability 21:01 - Funding updates: success with the Royal Academy of Engineering’s Green Future Fellowship 22:13 - Hardware refinement, engineering teams, and future milestones 23:15 - Focus on robotics applications and exploring broader use cases 24:54 - Funding plans, runway, and potential for co-funding from industry partners 28:13 - Insights from the ChipStart program and strategic mentorship 30:39 - Meet Robert Swann, chairman with a rich industry background 31:37 - Future ambitions: balancing technical depths with business growth 34:11 - Advice for researchers aspiring to start their own ventures 35:36 - The significance of team and market understanding in startup success 36:14 - Competitive landscape and uniqueness of RAYD’s approach 37:25 - Final gratitude and future outlook Resources & Links: * Royal Academy of Engineering Green Future Fellowship [https://www.raeng.org.uk/education-and-impact/awards/green-future-fellowship] * Imperial College London [https://www.imperial.ac.uk/] * PSI (Photonic Systems in Switzerland) [https://www.psi.ch/en] * Robert Swan - LinkedIn [https://www.linkedin.com/in/robert-swan/], Alpha Mosaic [https://alphamosaic.com/] The BritChips Podcast is proudly sponsored by Silicon Catalyst UK, Official Partner of the government-funded ChipStartUK semiconductor incubator. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit anttheantidote.substack.com [https://anttheantidote.substack.com?utm_medium=podcast&utm_campaign=CTA_1]

BritChips podcast: Deep Fabrication - Predicting Defects

In this episode, Yasir Noori shares his entrepreneurial journey from quantum computing research to founding Deep Fabrication, a startup developing AI tools to predict defects in semiconductor manufacturing. Discover how AI can transform chip fabrication processes, reduce costs, and accelerate innovation in the industry. Key Topics * The evolution from quantum computing to semiconductor fabrication, and why semiconductors remain a passion. * How Southampton’s advanced clean room facilities influenced Dr. Noori’s focus on deep tech. * The origin story of Deep Fabrication and the importance of AI in defect prediction during manufacturing. * Details of the company’s flagship tool, Deep Litho, designed to optimize lithography processes. * The strategic approach to building a team, including insights from a “speed dating” exercise with AI experts. * The process of market testing with pilot projects within UK and European foundries. * The funding journey, including grants from Innovate UK, angel investors, and future plans for growth. * The balance between academic research, entrepreneurship, and personal life. * Lessons learned, including negotiation strategies and valuable business books like “The Lean Startup” and “The Mom Test.” * Future ambitions: expanding the AI tools across the entire semiconductor manufacturing chain. Timestamps (approx) 00:00 - Introduction to Dr. Yasin Noori and Deep Fabrication 02:00 - From quantum computing to semiconductors: a strategic career shift 04:00 - Southampton’s world-class fabrication facilities and their influence 05:30 - Inspiration behind Deep Fabrication’s defect prediction AI solution 07:00 - The concept development of Deep Litho and its potential impact 09:00 - Building a cross-disciplinary team through networking events 11:00 - Organic growth of AI models and expanding into manufacturing processes 13:00 - Critical steps of semiconductor fabrication and lithography explained 15:30 - How AI can optimize and accelerate lithography process 18:00 - Pilot testing with UK companies and plans for scaling globally 20:00 - Business model: licensing software and expanding modules 22:00 - Funding journey: grants from Innovate UK and plans for investment 26:00 - Navigating the challenges of startup incubation and university spinouts 33:00 - Managing time between research, entrepreneurship, and personal life 35:30 - Advice for young entrepreneurs and lessons learned along the way 42:00 - The importance of negotiation and customer validation strategies 43:00 - Future plans: raising funds, expanding team, and scaling impact Resources & Links * The Lean Startup by Eric Ries [https://www.amazon.com/Lean-Startup-Entrepreneurs-Continuous-Innovation/dp/0307887898] * The Mom Test by Rob Fitzpatrick [https://www.amazon.com/Mom-Test-customers-business-everyone/dp/1492180742/ref=tmm_pap_swatch_0?_encoding=UTF8&dib_tag=se&dib=eyJ2IjoiMSJ9.1qi3coaogGjWxa0BvlPeP_dFnC79M5uooqmtpl48Q2DP_705dDZtv6mFqSzRXtcwpHmiAFTcHNDbaVDff4R1D1tvtqH6kwsfT-ba3LH0fFYqcJYP-Huec4xF60-i0Y067zG79WJsckvSmX-lpvEUdK34f9Wp_67B8ji3L0OpNP9vo7L_uTsU2-uCYgtkRo5_e2Kk2_K-JeVgXtnLiBZw_Te4MpGeu7xVQpbDeP4BfRM.3__CWW-Qk58p4jqE3HZVRDWdQO14gdX5Omj8PEwJ2kw&qid=1776089164&sr=1-2] * Innovate UK Funding [https://www.gov.uk/government/organisations/innovate-uk] * Southampton University Zeppler Institute [https://www.southampton.ac.uk/zeppler-institute] * Future Worlds Southampton [https://www.southampton.ac.uk/innovation/future-worlds.page] The BritChips Podcast is proudly sponsored by Silicon Catalyst UK, Official Partner of the government-funded ChipStartUK semiconductor incubator. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit anttheantidote.substack.com [https://anttheantidote.substack.com?utm_medium=podcast&utm_campaign=CTA_1]