Sign in to view Chris C.H.’s full profile
or
New to LinkedIn? Join now
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
Sign in to view Chris C.H.’s full profile
or
New to LinkedIn? Join now
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
Newmarket, England, United Kingdom
![]()
Sign in to view Chris C.H.’s full profile
or
New to LinkedIn? Join now
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
384 followers
379 connections
Sign in to view Chris C.H.’s full profile
or
New to LinkedIn? Join now
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
![]()
![]()
![]()
![]()
View mutual connections with Chris C.H.
or
New to LinkedIn? Join now
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
View mutual connections with Chris C.H.
or
New to LinkedIn? Join now
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
Sign in to view Chris C.H.’s full profile
or
New to LinkedIn? Join now
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
About
Activity
384 followers
-
Chris C.H. Lin reposted thisChris C.H. Lin reposted thisEveryone’s talking about AI like it’s some mythical thing, when in reality, it’s just change management with faster tools. I was back on stage at the UK Retail Innovation Council Leaders’ Summit, where I spoke about how so many retailers are diving into AI pilots without stopping to ask three basic questions: 👉 Intent: Why are we doing this? 👉 Integration: How will this fit into how we work day-to-day? 👉 Impact: What will it actually deliver? Instead, I’m hearing from a lot of CIOs and CTOs who are under pressure from boards saying, “Where’s our AI plan? Our competitors are miles ahead!” (but they’re usually not). That pressure often pushes teams into doing something fast rather than doing the right thing, launching pilots, proofs of concept, and experiments just to show progress. The result is dozens of disconnected projects that never move past the pilot stage, no plan for how they’ll integrate into daily operations, and no real business impact to show for it. AI isn’t different. It still comes down to clarity, adoption, and scale, the same fundamentals that drive any successful change. So here’s what I tell retail leaders to focus on: ✅ Do: Start with a real business problem, not a technology trend. Get the people who will use AI involved early, they’ll make or break adoption. Plan how insights or automation will plug into everyday workflows, not sit on the side. Measure success in business outcomes, not models or pilots. 🚫 Don’t: Rush to deploy AI just to tick a box. Let every department run its own experiment without coordination. Assume that “having data” is the same as being data-driven. Underestimate the cultural change required for teams to trust and use AI outputs. The best AI strategies look a lot like the best change strategies; they’re intentional, integrated, and focused on real impact.
-
Chris C.H. Lin reposted thisChris C.H. Lin reposted thisManual processes. Disconnected data. Slow answers for employees and consumers. Here’s how one travel company is using AI to fix it and build a smarter way to work. https://lnkd.in/ez95MYsM #IgniteAIPartners #AIInConsumerServices #AIInTravel
-
Chris C.H. Lin reposted thisChris C.H. Lin reposted thisSay hello to Yaraku Translate — your smarter way to translate. Built on AI + human collaboration, our next-generation CAT tool just got a new name to match its mission. Formerly known as Yarakuzen, our platform has been officially rebranded as Yaraku Translate as of April 2025. The new name better reflects our commitment to making high-quality, intuitive translation accessible to all. Since its launch in 2014, Yaraku has helped over 2,000 companies streamline multilingual communication with secure, easy-to-use AI translation technology. This rebranding marks a new chapter — but our mission remains the same: empowering global teams through smarter translation. 🔗 Learn more about the rebrand: https://www.yaraku.com/ #YarakuTranslate #Translation #AI #MachineTranslation #Localization #Productivity #Rebranding
-
Chris C.H. Lin reposted thisChris C.H. Lin reposted thisBig conversations, big ideas. Where does AI fit in your business? Yesterday at the Institute of Directors (IoD)’s Better Director Series in Leicester, we had the chance to hear from incredible speakers on HR, finance, and business strategy, giving leaders practical takeaways to shape their future success. Our Co-Founder, Craig Bentley, joined the discussion to talk about AI’s role in smarter decision-making, and it’s clear that businesses are still figuring out where it fits. One of the biggest mistakes? Starting with the question: “Where should I put AI?” Instead, a smarter approach is: What are you trying to achieve? Are you trying to cut costs? Do you want to improve customer experience? Is it a quality or efficiency issue? Once you know what you’re trying to achieve, AI can help get you there. But it’s about strategy, not just technology. Of course, AI isn’t without risks. We talked about: Why human validation is crucial. AI shouldn’t replace human decision-making entirely. The need for AI guardrails to avoid costly mistakes. Why businesses must educate teams on risks like exposing proprietary data to public models. We’re just at the start of AI’s journey, and things are moving fast. From AI agents that can make decisions autonomously to intelligent automation transforming industries, this is a once-in-a-generation shift. 💡 Where are you with your AI readiness? If you’re wondering how AI fits into your business, let's talk. Huge thanks to Dr Debra Charles (Hons), CEO, Novacroft, John Browett, Chair, Institute of Directors & Octopus Group, and Lewis Stringer, Senior Manager, British Business Bank for the brilliant discussion. #IgniteAIPartners #AIReadiness
-
Chris C.H. Lin reposted thisChris C.H. Lin reposted thisBREAKING NEWS The Royal Swedish Academy of Sciences has decided to award the 2024 #NobelPrize in Physics to John J. Hopfield and Geoffrey E. Hinton “for foundational discoveries and inventions that enable machine learning with artificial neural networks.” This year’s two Nobel Prize laureates in physics have used tools from physics to develop methods that are the foundation of today’s powerful machine learning. John Hopfield created an associative memory that can store and reconstruct images and other types of patterns in data. Geoffrey Hinton invented a method that can autonomously find properties in data, and so perform tasks such as identifying specific elements in pictures. When we talk about artificial intelligence, we often mean machine learning using artificial neural networks. This technology was originally inspired by the structure of the brain. In an artificial neural network, the brain’s neurons are represented by nodes that have different values. These nodes influence each other through connections that can be likened to synapses and which can be made stronger or weaker. The network is trained, for example by developing stronger connections between nodes with simultaneously high values. This year’s laureates have conducted important work with artificial neural networks from the 1980s onward. John Hopfield invented a network that uses a method for saving and recreating patterns. We can imagine the nodes as pixels. The Hopfield network utilises physics that describes a material’s characteristics due to its atomic spin – a property that makes each atom a tiny magnet. The network as a whole is described in a manner equivalent to the energy in the spin system found in physics, and is trained by finding values for the connections between the nodes so that the saved images have low energy. When the Hopfield network is fed a distorted or incomplete image, it methodically works through the nodes and updates their values so the network’s energy falls. The network thus works stepwise to find the saved image that is most like the imperfect one it was fed with. Geoffrey Hinton used the Hopfield network as the foundation for a new network that uses a different method: the Boltzmann machine. This can learn to recognise characteristic elements in a given type of data. Hinton used tools from statistical physics, the science of systems built from many similar components. The machine is trained by feeding it examples that are very likely to arise when the machine is run. The Boltzmann machine can be used to classify images or create new examples of the type of pattern on which it was trained. Hinton has built upon this work, helping initiate the current explosive development of machine learning. Learn more Press release: https://bit.ly/4gCTwm9 Popular information: https://bit.ly/3Bnhr9d Advanced information: https://bit.ly/3TKk1MM
-
Chris C.H. Lin reposted thisChris C.H. Lin reposted thisBuilding a multi-agent financial concierge 💸🤖 This is one of the cleanest implementations of a complex multi-agent system that I've seen. Props to Laurie Voss. All the "multi-agent" code can be implemented in a single Python class as a set of decomposable steps. You get ✅ all the benefits of an event-driven architecture (high scalability, visualizations, real-time) ✅ all the customizabillity of building from scratch with our core LLM/prompt modules ✅ the cleanliness of writing orchestration directly as Python code instead of baking it into graphs. End-result: a collaborative multi-agent system powering a customer support bot in a bank that allows the user to look up information, authenticate, and perform "actions" within their account once authenticated. Repo: https://lnkd.in/gWucYuCm Video: https://lnkd.in/gzV2XqhT
-
Chris C.H. Lin reposted thisChris C.H. Lin reposted thisLinkedIn has published one of the best reports I’ve read on deploying LLM applications: what worked and what didn’t. 1. Structured outputs They chose YAML over JSON as the output format because YAML uses less output tokens. Initially, only 90% of the outputs are correctly formatted YAML. They used re-prompting (asking the model to fix its YAML responses), which increased the number of API calls significantly. They then analyzed the common formatting errors, added those hints to the original prompt, and wrote an error fixing script. This reduced their errors to 0.01%. 2. Sacrificing throughput for latency Originally, they focused on TTFT (Time To First Token), but realized that TBT (Time Between Token) hurt them a lot more, especially with Chain-of-Thought queries where users don’t see the intermediate outputs. They found that TTFT and TBT inversely correlate with TPS (Tokens per Second). To achieve good TTFT and TBT, they had to sacrifice TPS. 3. Automatic evaluation is hard One core challenge of evaluation is coming up with a guideline on what a good response is. For example, for skill fit assessment, the response: “You’re not a good fit for this job” can be correct, but not helpful. Originally, evaluation was ad-hoc. Everyone could chime in. That didn’t work. They then have linguists build tooling and processes to standardize annotation, evaluating up to 500 daily conversations and these manual annotations guide their iteration. Their next goal is to get automatic evaluation, but it’s not easy. 4. Initial success with LLMs can be misleading It took them 1 month to achieve 80% of the experience they wanted, and additional 4 months to surpass 95%. The initial success made them underestimate how challenging it is to improve the product, especially dealing with hallucinations. They found it discouraging how slow it was to achieve each subsequent 1% gain. #aiengineering #llms #aiapplication
-
Chris C.H. Lin reposted thisChris C.H. Lin reposted thisSakana AIの最初の研究成果である、進化的計算による基盤モデル構築に関する論文を公開しました。多様な既存モデルを自動的に融合し優れた基盤モデルを構築するための方法を提案すると共に、それにより試作したモデルを公開しました。 ブログ https://lnkd.in/gin7mWpi 論文 https://lnkd.in/gZacPnRu 我々の開発した手法「進化的モデルマージ」は多様な能力を持つ幅広い既存のモデルを融合して新たな基盤モデルを構築するための方法を進化的アルゴリズムを用いて発見する手法です。既存のオープンモデルの膨大な集合知を活用出来るため、モデルを非常に効率的に作成できます。私たちのアプローチは、「非英語言語と数学的推論」「非英語言語と画像」といった、これまでは困難と思われていた全く異なる領域のモデルを融合する方法すらも自動的に発見できることが分かりました。 この手法を実証するため、3つの日本語基盤モデルを試験的に構築しました。日本語大規模言語モデル「EvoLLM-JP」、日本語で対話可能な画像言語モデル「EvoVLM-JP」、高速な日本語画像生成モデル「EvoSDXL-JP」、それぞれのモデルが客観的評価において非常に高い性能を達成しています。現在EvoLLM-JPとEvoVLM-JPのモデルを公開している他、EvoVLM-JPをご自身の画像でお試し頂けるデモを公開しています。 GitHub https://lnkd.in/gNKWbuvV デモ https://lnkd.in/g6BmB5ah
-
Chris C.H. Lin liked thisChris C.H. Lin liked thisVery cool to be back at University of Warwick - Warwick Business School 👨🏽🎓 20 years Ross Ritchie since we started our MBA here… and we haven’t changed a bit 🙃
-
Chris C.H. Lin liked thisChris C.H. Lin liked thisThis photo was taken on June 2, 2023; roughly two months later, Yang Zhilin founded Moonshot AI. The world was changed forever.
-
Chris C.H. Lin liked thisChris C.H. Lin liked thisMore highlights from the Cerebral Valley AI Summit: - Recursive's Josh Tobin believes AI research is on the cusp of the same shift deep learning brought to hand-crafted features - this time with models doing the research. Recursive self-improvement arrives when the primary innovations in a new model come from the previous version of the model itself. On how a startup competes with frontier labs: ability to focus, while big-lab compute is spread thin. - Sequoia Capital's Luciana Lixandru sees application-layer companies like Harvey and Sierra crushing it, but the next cohort is less obvious — the question is your right to win, and the answer is domain expertise. On defence, sees Stark as a platform company and the West has been underinvesting. Cautiously optimistic on Europe: the ecosystem is finally producing repeat founders. - Index Ventures' Danny Rimer thinks it's too early to call which foundation models will stick around - "we're in the early innings". Differentiation comes from the approach taken, the talent it attracts, and the culture that keeps them, and he was notably effusive on Anthropic's clarity in what they stand for. On hiring he cited Notion's Ivan Zhao: evaluate on capability, taste, and agency - credentials matter less than ever.
-
Chris C.H. Lin liked thisChris C.H. Lin liked thisLloyd's of London started in a coffee house so we figured the market was overdue dessert... Yesterday afternoon we parked an ice cream van next to the Lloyd's building and gave out free ice creams to anyone who works in insurance. Over 400 people came by in three hours! The best part was meeting customers we'd only ever seen on a call, and talking to so many people in the market we'd never met before. Thanks to everyone who came by, and to the team for pulling it off 🍦
-
Chris C.H. Lin reacted on thisChris C.H. Lin reacted on thisWe’re outside the Lloyd’s building in London today with free ice cream for insurance workers to help you recover from last night’s game 🍦 Come and grab an ice cream, we’re here until we run out! ☀️ Jenny Hu Jack Fox Laura Grainger Nasri EL-Sayegh Hannah Smallman Sasha Haco Laura Hauser Ben Hounsell
-
Chris C.H. Lin liked thisChris C.H. Lin liked thisSpent last weekend at the Poolside Laguna XS.2 Hackathon, run together with Prime Intellect, NVIDIA and Hugging Face. A refreshing dip back into the lagoon of fine-tuning open-source models, rather than just building on top of LLM APIs, and a good reminder that this may be more of the future than it currently feels. My team's project: a Protein-Ligand Design Gym to teach Laguna XS.2 the chemistry of how small molecules bind to proteins. The trick is verifiable rewards. Every question has a single checkable answer the model can only reach by calling real chemistry and biology tools, the way a scientist would. The hackathon's over but the work isn't. The Gym now has 1,000+ questions, and the full dataset is open on Hugging Face for anyone who wants to build on it. Next up: adapting the Laguna architecture to take proteins and molecules as direct input, along the lines of the recent Proteo-R1 work on reasoning models for protein design. Around 30 teams turned up and the bar was high. Full-duplex dialogue modeling, CUDA kernel distillation, KV-cache compression, native vision encoders. Congrats to the winners, and thanks to poolside for running it. https://lnkd.in/eR6bTw5Z
-
Chris C.H. Lin liked thisChris C.H. Lin liked thisHi everyone, I want to share that today marks the end of my journey with Citrix. After almost 27 years contributing to the growth of the Citrix and NetScaler businesses—across 9 different roles in 3 countries—it is time to close this chapter of my professional career. It has been an absolute privilege to learn and grow alongside so many incredible colleagues, partners, and customers over the years. I am deeply grateful for the mentors who guided me, the executive leadership teams who trusted my vision, the exceptionally talented cross-functional teams I collaborated with across our global offices, and the humbling opportunity to lead such remarkable people across my 25 years in leadership roles. I want to specially acknowledge those who made it possible for my family and me to move to Madrid, Spain. It gave us the opportunity to build new roots and be surrounded by a wonderful community of people who have made both my professional and personal life incredibly fulfilling and memorable. There are simply too many people who have positively impacted my development to name everyone individually. However, I truly look forward to staying in touch with the many of you with whom I have shared a special connection throughout this journey. I wish you all the very best. May your work always bring you closer to your biggest goals in life, and may you continue to thrive both professionally and personally. As this chapter closes, a new one is about to start—brighter and wiser. Let's stay connected! All the best, Jair #Citrix
-
Chris C.H. Lin liked thisChris C.H. Lin liked this🥵 An intense heatwave in Europe has shattered numerous temperature records and had major impacts on human health, ecosystems, agriculture, infrastructure and labour productivity. Millions of people are now under the maximum warning level - a Red Alert - for extreme heat, as the World Meteorological Organization, its members and partners mobilize with early warnings and coordinated heat-health action plans to try to save lives and support emergency planning. Here's a snapshot: France: Hottest day on record on 24 June, with average national temperature of 30.0° C, per METEO FRANCE. Temperatures rose up to 43.8 °C in the town of Pulluau in western France. Overnight temperatures also set a new national record of 22 °C. Spain recorded its hottest June days on record on 23 and 24 June. The United Kingdom broke the June temperature record for three consecutive days, with 36.4 °C recorded in southern England on 25 June. The Met Office issued Red Warnings for extreme heat for three straight days - the first time it has done this. Germany is under widespread extreme heat alerts, with temperatures forecast to increase this weekend, says the Deutscher Wetterdienst. The KNMI - Royal Netherlands Meteorological Institute issued an unprecedented Red Alert for extreme heat for eight provinces for 26 June. Switzerland set a new June temperature record of 38°C in the northern city of Basel, according to Federal Office of Meteorology and Climatology MeteoSwiss. This is the extraordinary heat seen from space on 24 June by the #Sentinel3 satellite of the #Copernicus program via adamplatform. Note that these are land surface temperatures which are higher than air temperatures. Read the WMO news story: https://lnkd.in/dHz6bfDE
-
Chris C.H. Lin liked thisI look forward to presenting at this event!Chris C.H. Lin liked thisRobotics adoption is no longer a future consideration for many SMEs; it's becoming a business necessity. But what does that look like for you? Join us on the 14th of July at the Robotics Adoption Forum and meet with industry experts, robotics innovators, technology leaders, and fellow SMEs to explore the opportunities, challenges, and practical realities of robotics. You will also hear from Extend Robotics, as they share insights from their growth journey with practical advice for businesses looking to adopt robotics. If you're lucky, you might even get to meet some of their robots 🤖 Registrations are now open, but places are limited -> https://lnkd.in/ejYPyRAy
Experience & Education
-
Unitary
******** ********
-
****** ** ********
**** ** *********** ********
-
*******
******** ******** * **** *********
-
*** ***** ********* ** **********
****** ** ******* * ** ******* ******** GPA: 5/5
-
-
*** ******* ********** ** **** ****
******** ** ******* * ** ******* ***** ******** ***** *******
-
View Chris C.H.’s full experience
See their title, tenure and more.
or
Licenses & Certifications
Languages
-
Mandarin Chinese
Native or bilingual proficiency
-
English
Native or bilingual proficiency
-
Japanese
Limited working proficiency
-
Swedish
Elementary proficiency
-
Cantonese
Native or bilingual proficiency
Other similar profiles
Explore more posts
-
The Alan Turing Institute
🔎 Unlocking reasoning and planning capabilities within LLMs This week Turing researcher, Pranava Madhyastha, presented the latest results from our work to improve the reliability and consistency of large language models (LLMs). Using techniques from the field of symbolic AI, Pranava and team are developing a framework that applies constraints to LLM outputs, helping to reduce the amount of false, incoherent or inconsistent content that they generate. This work aims to improve the reasoning and planning capabilities of LLMs, which will make them more suited to deployment in safety-critical areas such as healthcare and engineering. ➡️ Read more about the project: https://bit.ly/4iwJ9lv
76
2 Comments -
PyTorch
Our latest PyTorch community blog from the SGLang Team covers how SGLang advances hybrid Mamba plus Attention models with improved memory efficiency, prefix caching, speculative decoding, and serving performance. 📖 Read the blog: https://lnkd.in/gfDZkZWF #PyTorch #AIInfrastructure #SGLang #LLM
112
1 Comment -
Byte Goose AI
Transformers and Large Language Models Cheatsheet The podcast provides a comprehensive overview of Transformer architectures and Large Language Models (LLMs), detailing their fundamental mechanics, training lifecycle, and practical applications. The materials explain core concepts such as tokenization, embeddings, and the Multi-Head Attention (MHA) mechanism, clarifying the distinct roles of the original Encoder-Decoder architecture and its specialized variants, Encoder-Only (BERT) and Decoder-Only (GPT) models. Furthermore, the podcast outline the three main stages of the LLM lifecycle—pretraining, finetuning, and preference tuning—and introduce advanced techniques like Retrieval-Augmented Generation (RAG), LLM Agents (ReAct), and computational optimizations such as Flash Attention and Parameter-Efficient FineTuning (PEFT). The content is presented across study guides, whitepaper excerpts, and cheatsheets, covering everything from architectural components to practical deployment strategies. https://lnkd.in/gESJCZTf
2
-
Quantum Zeitgeist
Guidelines Advance Large Language Models for Superior Code Generation Researchers have identified ten practical guidelines , focusing on clear input/output specifications, detailed examples, and ambiguity reduction , that demonstrably improve the effectiveness of large language models for code generation, even when developers initially underestimate their value. #quantum #quantumcomputing #technology https://lnkd.in/eJJF6DCE
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content