Bryan Tom

This week, I made it back to MWC after a few years away. It was amazing to see the hustle, the booths, and more importantly to reconnect with friends, colleagues and partners in the Fira (and the "after-hours Fira" (aka Temple)). SoftBank also unveiled its Telco AI Cloud vision (https://lnkd.in/gzQ7SyR2) . On the surface, it looks like another GPU cloud play. But look closer, and you'll see something fundamentally different: a telecom operator rewiring itself into an AI-first operator. Beyond deploying GPUs in bulk, its about delivering a simplified but powerful platform to enable AI-on-RAN and eventually AI-for-RAN use cases. A key component (and a genuinely hard problem) is resource and workload management. When you're running AI workloads across centralized data centers AND edge MEC nodes AND live RAN infrastructure simultaneously, intelligently managing hardware, containerized workloads and ensuring the right workload runs in the right place while meeting expected quality and performance levels becomes non-negotiable. Failures will be noticed. That's where Infrinia AI Cloud OS comes in. It is the software stack that ties together compute resources and workloads, and supports orchestrating workloads across the entire environment. The goal: compute that flows fluidly between training, inference, and network control based on actual demand, across core, edge and deep edge. Telcos have always owned the physical layer. Having spectrum and proximity to users is an incredible advantage and with the right software stack, that becomes a strategic moat in the AI era. Grateful to be part of this journey! https://lnkd.in/gzQ7SyR2 #airan #infrinia #physicalAI #TelcoAICloud

38

1 Comment

How did LangTalks end up meeting with Andrew Ng, Anshul Ramachandran (Windsurf), and Roey Zalta And what book does #LarryPage wish he had when he started Google? It all relates to the question that keeps me up at night: how should our evolving understanding of 'search & retrieval', 'context engineering', and 'evals' shape the way we design systems? And how should we account for these shifting dynamics in the planning stage with a forward looking approach? (Though Page’s recommendation is real, so I hope you read until the end.) First, the new course Agentic AI with Andrew Ng is one of the best resources I’ve found (especially Module 4 on evals) [1]. I also recommend listening to LangTalk’s episode on search & retrieval with Guy Itach [2]. I especially liked the ending, where they show how technical aspects are combined with product sense and domain expertise, and why choosing the right evals isn’t straightforward at all. Don’t miss Roy Zalta’s post [3] on how he achieved 98% accuracy in his RAG system using a combination of eleven different strategies. And last but not least: Windsurf’s AI Coding Agents [4]. Specifically, the section on “Search & Discovery for AI Agents,” where Anshul presents a multi-step retrieval paradigm—that blew my mind (it’s free). Another thought I’ve had on 'evals and error analysis' is that it's somewhat derivative of the OKRs methodology—an incredibly powerful concept for driving clarity, prioritization, organizational alignment, and decision-making. I honestly can’t recommend Measure What Matters [5] enough—but don’t take my word for it. See what Larry Page, Alphabet CEO and Google cofounder, had to say: "I wish I had had this book nineteen years ago, when we founded Google. Or even before that, when I was only managing myself! As much as I hate process, good ideas with great execution are how you make magic. And that’s where OKRs come in." Circling back to my question: As we shift toward agentic systems, how do you see the evolution of system design? What planning strategies are you starting to prioritize from the very beginning?

6

1 Comment

🚀 Reimagining Product Discovery at Scale As part of our 2025 reset, we ran dual-location discovery workshops for our stakeholders situated across 🇮🇩🇬🇧🇸🇬🇲🇾🇮🇳 to align cross-functional teams around a shared vision. Our prioritization lens🔎 • Impact/Urgency matrix ⚓️ • Strategic themes (time-boxed bets) ⚡️ • OKR alignment 📊 • Client urgency & revenue impact 💰 • Voice of customer across Sales, Ops, CS, and Partners 🏬 What this delivered: 💪🏻 ✅ Data-driven prioritization with an impact-vs-cost view ✅ Stronger alignment across hubs and functions ✅ A clear 2025 roadmap shaped with Engineering from day one ✅ Tighter collaboration rhythms for Product × Tech × GTM If you’re scaling discovery across locations and want to compare notes, happy to share playbook*. Comment - "Discovery", I will be happy to share it #productleadership #productdiscovery #roadmapping #productstrategy #crossfunctional #b2b #saas #orgdesign #uxresearch #engineering #innovation (*Views are my own. High-level methods only; no confidential details.)

19

4 Comments

I just finished watching all of the top AI-related sessions from the World Economic Forum 2026, with leaders like Jensen Huang, Satya Nadella, and Dario Amodei. The consensus? The constraints that restricted our abilities to drive product innovations years ago have largely evaporated. Here are 5 "Impossible" Product Ideas from two years ago that are now engineering realities: 1. The "Self-Coding" Feature Suite 💻 • The 2026 Reality: The loop is closing. Anthropic predicts that within months, models will perform most maybe all of what software engineers do end-to-end. We are moving from "Human in the loop" to "Human in the lead," meaning products can now autonomously maintain and upgrade themselves. 2. The "Universal Shopper" That Actually Buys 🛒 • The 2026 Reality: "Agentic Commerce" is here. Visa has rolled out protocols where agents carry attribute payloads (spending limits, preferences) to buy natively on platforms. The buy button is moving from the merchant's site directly into the AI agent. 3. The "Physical" Health & Care Assistant 🏥 • The 2026 Reality: AI has moved to "Physical Intelligence." It now understands the laws of physics and biology. With humanoid robots slated for public sale by next year, the hardware finally exists to deploy the software we dreamed of. 4. The "Brute Force" Intelligence Engine 🧠 • The 2026 Reality: The cost of intelligence has collapsed. Inference costs dropped 99% from GPT-4 to GPT-5 mini. We can now afford to throw massive test time compute at problems, allowing the model to think for minutes before answering, drastically reducing hallucination. 5. The "Sovereign" Enterprise Brain 🔐 • The 2026 Reality: Sovereignty is the new standard. The focus has shifted to "Sovereignty of the Firm", distilling your own private data into a model you control, ensuring your IP doesn't leak into a generic API. The agents are ready to work. Time to find that dry-erase marker!

2

Many founders and tech professionals are starting to explore the O-1 visa, often assuming that building a strong O-1 profile can be done quickly. However, based on my conversations with Jason Cheung about EB-1A and O-1 profile-building, this is not the case. Even O-1 profile-building can take months or even several years. If you want to start building an O-1 profile, begin by: - Getting involved in activities relevant to the criteria you want to target. - Speaking to reporters if you are already connected to them through your network or company. - Attending conferences to speak, which can help demonstrate your expertise in your field. To be notified first-hand about all our future opportunities like these, subscribe to our calendar of events: https://lu.ma/ventureminer Additionally, if you're interested in learning more about EB-1A and O-1 profile-building, check out Jason's profile-building opportunities database, private group, and other resources through his newsletter (only $12 per month): https://lnkd.in/deqhfpRs #eb1a #o1visa #greencard #swe #tech

13

Building Large Language Models for the Public Sector on AWS Niki Sotiria Kokkalas Eliuth Triana Laura Verghote Wenhan Tan Excited to share our latest deep-dive on developing custom LLMs tailored for government and public sector organizations! 🏛️🤖 Off-the-shelf models often fall short of the regulatory, cultural, and operational requirements that define public sector missions. This comprehensive guide walks through the complete development lifecycle: ✅ Use case & requirements definition ✅ Evaluation framework establishment ✅ Model selection & training approach ✅ Data collection & preparation ✅ Infrastructure provisioning ✅ Production deployment & testing Key highlights: 🌍 National LLMs that preserve linguistic heritage & cultural context 🔒 Data sovereignty & compliance frameworks 💰 Cost analysis: API vs. self-hosting TCO 🛠️ Tools: NeMo Curator, SageMaker HyperPod, LM Harness Whether you're building language models for underrepresented languages or creating domain-specific solutions for healthcare, education, or legal services—this guide provides the scientific methodology and AWS infrastructure blueprint you need. Read the full article: https://lnkd.in/eKZ8FaKN Would love to hear your thoughts and experiences with custom LLM development! 💬

166

10 Comments

Last week, Signifyd’s AI Lab, Data Science, and Engineering teams hosted Dr. Hui Wang for a fireside chat. It is rare to find a leader who can seamlessly pivot from the theoretical "Innovation Paradox" of LLMs to the practical challenges of leading through a period of rapid change. She provided a roadmap that balanced radical transparency with clear executive boundaries, covering an incredible range of ground in just under an hour: * The Productivity Paradox: AI can speed up coding by 50%, yet corporate impact often stays in the single digits. Why? Because the bottleneck often isn’t the tech, it’s the "human maneuvering." Without bridging the gap between decision-makers and data, the code can’t save us. * The "Default" Trap: When data is limited, the "truth" is often hidden. Without intentionality, we will automate the invisibility of those who don't fit the mold. That is where human intuition becomes a technical requirement. * Sponsorship > Mentorship: True growth doesn't just happen through skill-building; it happens through sponsorship, when a leader mentions your name in a room you aren't in yet. * Controlled Imbalance: Rejecting the myth of "work-life balance," Hui shared her practice of "controlled imbalance" - hard-blocking 5 PM – 9 PM for family and being unapologetic about it. Special thanks to Westin Sherlock Jing Xue Robert King and all those who contributed to a lively discussion! Spending this time with Dr. Hui anchored our Women’s History Month celebrations in a way that felt both timely and profoundly authentic, for all of use here at Signifyd. Thank you so much, Hui!

98

3 Comments

#amazonconnect #CCaaS #customersuccess 🏛️ The Wisconsin Department of Revenue (DOR)'s Contact Center, which handles 700,000 annual calls, faced challenges with outdated technology, frequent outages, and inefficient workflows. Partnering with AWS Partner ScaleCapacity, DOR successfully modernized its technology by migrating to Amazon Connect in just four months. This transformation improved both customer and agent experiences while dramatically reducing operational costs. Results 🎉: 🌟 66% reduction in technology costs 🎯 Complete elimination of system outages ⚡ 60% decrease in hold times (from 1:44 to just 42 seconds) Features They Valued: 🌟 • AI summaries & transcripts 🤖 • Real-time analytics 📊 • Auto case creation ⚙️ • Smart routing ⚡ • Quality management 💰 • Pay-as-you-go pricing Final Result: A super-reliable system with happier customers and staff! Read more here->

15

I wrote an article on how we translate user-generated content at scale at Mercari. It summarizes two years of work with various models, LLMs but not only, and how cost evolved, reducing it 100x over the period. I also go over non-AI features and user experience. https://lnkd.in/gbpQAQmX

39

2 Comments

Closing the week by celebrating another great customer story — this time from Kakao Corp In this latest blog, Kakao shares how they overcame infrastructure scaling bottlenecks by moving to JAX + TPUs, enabling faster training of their Kanana model family. Their goal was to build more efficient, adaptable AI systems at production scale. With the right hardware + software combination — JAX for flexibility and TPUs for performance — customers can unlock breakthrough speed, efficiency, and scale in their AI workloads. Excited to help bring these innovations to organizations worldwide. https://lnkd.in/gu8UAk57 #AIleadership #Jax #TPUs #ScalableAI

10

Had such a great chat with Vivien Lin! She brought so much wisdom and vivid insights to the conversation: 1. We're trying to work with our users, when they make money we make money. We try not to stand on the opposite. 2. From the start we decided to exit China market. We never touched US users. It's all because we are quite self-disciplined and self-regulated. 3. We consider whether this product or whether a feature can truly solve the problem for the user. 4. From 2018, 2019 is always put our top priority to think about how we bring more new user to the crypto industry. It's not like competing the existing user with other exchange. 5. Ideally, the final product should be you don't need to worry about what functions sit in what page. You just go to the AI or the AI assistant to tell him what you want. 6. The difference in the roadmap between crypto and traditional finance is rooted in how the industry got originated. For traditional finance, banking and payment that's naturally where finance comes from than lending than trading. At crypto, the very first need where the industry originated is mining. 7. I think, we are still in a very precious time window especially for exchange it to become a sort of mega platform in near future. 8. My suggestion is always be open minded - open minded to the technology, business model and even all those new ideas. 9. My suggestion to everyone, even if you are a legal, you are a product manager or you are in the operation team, is to study programming, learn how to code. 10. The slides always tell the good stories but you need to get down to the details to understand what's happening. 11. Before all those regulations put into place there has to be someone familiar with how possible the regulation is going to be implemented in the future and we can get prepared beforehand. 12. Without one major hack exchange cannot be a mature exchange. 13. If I know what they are doing I can read their code, I can get a very direct sense about the quality of the project instead of really looking at the powerpoint. Watch the full session here: https://lnkd.in/dGhbXdwr BingX

10

Welcome to the Episode 4 of building a product live on LinkedIn!! This is the funny part of building in public, it took me longer to pen down my journey on Medium ( https://lnkd.in/gJ3MZDkf ) than to actually do the work. (And unironically, exactly the problem we are trying to solve for PMs) Quick recap if you’re new here: I’m building Atlas, an AI copilot for product managers, live. In the first three posts, I focused on the 0 to 1 PM basics: figuring out the real problem, then setting goals and KPIs that actually mean something. Now I’m in my favorite messy zone: turning all of that into UX workflows and wireframes. Instead of “AI chat for PMs,” I tried to design around real jobs a PM wakes up with: • “Catch me up” • “Give me a first PRD draft” • “Show me what users are really saying” • “Help me define success” A few rules I kept repeating to myself: • Design workflows, not clever features • Make inputs and outputs painfully explicit • Bake in traceability from day one • Let one workflow feed the next After drafting the workflows with care, I turned the specs into action. I asked ChatGPT to generate a precise prompt tailored for a no-code wireframe builder. I then parsed that prompt into Base 44 (one of several viable tools) and produced a dynamic, clickable wireframe. If you want to play around with a fully working high fidelity wireframe, don’t hesitate to reach out. Stay tuned for part 5 as the journey has just started to pick up the pace! #AIProducts #ProductManagement #VibeCoding

15