FPL Technologies

Everyone is watching the S&P 500 and Nasdaq. But one of the biggest market rallies this year is happening in South Korea. The KOSPI, South Korea’s leading index is up 80%+ in just four months of 2026, and almost 3x in over a year, all driven by the AI boom. Why? The KOSPI is heavily concentrated in tech and semiconductors, with ~50% of the index tied to companies like Samsung Electronics and SK Hynix. KOSPI also has other heavyweights like Hyundai, Kia, LG, but the market is right now all driven by chip based companies. Driven by an AI-powered chip rally, Samsung's market capitalization surpassed $1 trillion, with Samsung and SK Hynix together accounting for over 43% of the KOSPI's total value. What makes this rally even more interesting is that for most of the last 5 years before 2024, the KOSPI delivered muted/negative returns. #KOSPI #AI #SAMSUNG

What the f**k is even a p-value? 🤯 A/B testing is one of the most important skills in Product, Growth, and Analytics... but most people drop it because stats feels unnecessarily complicated. P-values. Statistical significance. Confidence intervals. The concepts are simple. The jargon makes them scary. So I built a free interactive A/B Testing Guide that explains everything like you're a kid. 🍕 Pizza shop analogies for Control vs Treatment 🕵️ Detective-style hypothesis testing 🧮 Interactive calculators for significance, SE, and power No boring textbook stuff. Just practical understanding. If you work in Product, Analytics, Growth, or Startups, this is a career superpower. 🔗 https://lnkd.in/drFGZFuE If this helps someone in your network, repost it ♻️ #ProductManagement #DataAnalytics #ABTesting #Growth #ProductAnalytics #Experimentation #Product #Analytics

Product Builder role is here — ft. Razorpay Level 1: Use AI to generate primary flows and fallback scenarios. Level 2: Use AI for rapid prototyping. Level 3: Start shipping products directly in code. Level 4: Build specialized skills — best frontend practices, generate designs for different personas, apply UX principles. Ship locally. Stress-test edge cases. Validate early drafts with real users. Create and connect APIs. Fix what’s broken. Push to production. Go live. Razorpay Saurabh Soni

This is my first LinkedIn post, and honestly, it took a "system reboot" to get me here. After 6 years in the industry, I decided to jump into Shravan Tickoo ’s AI Product Manager cohort. No certificates, no fluff - just raw building and late-night logic. What did 9 weeks of 5-hour weekends actually look like? • Shipping with AI: Moving past the hype to understand real AI capabilities. Thanks to Attharv Sardesai , I even found myself unlocking Claude via the terminal-something I didn't see coming! • The 360° PM View: Learning to look at problems through UI, UX, and Backend lenses simultaneously, always anchoring back to First Principles. • The "Thala" Grind: Shoutout to Team Thala ( Krishna Akhil Allumolu Aditya Jiddu Rahul Kumar Vibudh Vishal Deepti Sharma Preeyashree Mallick . Our daily 9 PM to 1/2 AM calls were where the real magic happened. We didn't just meet; we pushed each other to deliver on time while actually hitting the mark. • The Clarity: Shravan’s sessions on the Bhagavad Gita were a game-changer. They taught me how to view problems with clarity and stop "creeping" on the small stuff. A huge thanks to Unnati Nakra and Gnanesh L for keeping me moving in the right direction, teaching how to document things like a pro, and ensuring thinking remained sharp and first-principle-led. I'm walking away with an "AI side" I didn't know I had and a squad that proved "hard things" are better handled together. If you’re facing a problem today, try looking at it from the first principle. It usually works. 🛠️ #ProductManagement #AI #FirstPrinciples #TeamThala #RethinkSystems #BuildingInPublic #FirstPost

𝐃𝐀𝐘 𝟏𝟏: 𝐀𝐈 𝐄𝐯𝐚𝐥𝐬 Hot take: most teams ship AI features with no real idea if they're working. 📊 Not "did users click it." I mean, is the AI producing good outputs? Reliably? Consistently? At scale? That's what #AIEvals are for. Three types every PM should know: 𝐂𝐨𝐫𝐫𝐞𝐜𝐭𝐧𝐞𝐬𝐬: Is the output factually accurate? Did the AI hallucinate? 𝐐𝐮𝐚𝐥𝐢𝐭𝐲: Is it well-structured, appropriately detailed, on-brand? 𝐁𝐞𝐡𝐚𝐯𝐢𝐨𝐮𝐫: Did it follow instructions? Stay within guardrails? Use tools correctly? Here's why evals are a PM problem, not just an engineering one: You wouldn't ship a payments feature without testing every transaction edge case. "Relying only on user feedback" is not a testing framework. 🎯 Without evals, you can't improve systematically. You're just guessing. Start simple. Define what "good" looks like. Build test cases. Score outputs. Find patterns in failures. Fix them. Repeat. That loop is your AI quality roadmap. Tomorrow: what if you want to change how the model itself thinks and not just what it's told? 👇 #ResponsibleAI #GenerativeAI #AIPM #AIAccuracy

𝐃𝐀𝐘 𝟏𝟎: 𝐌𝐂𝐏 You've built an AI agent. It's smart. It can draft emails, schedule meetings, update your CRM. Brilliant. One problem: It has no access to anything. It's like hiring the best assistant in the world and giving them no laptop, no phone, no access to any system. The intelligence is there. The hands aren't. 🙌 This is exactly the problem #MCP solves. 𝐌𝐂𝐏 = 𝐌𝐨𝐝𝐞𝐥 𝐂𝐨𝐧𝐭𝐞𝐱𝐭 𝐏𝐫𝐨𝐭𝐨𝐜𝐨𝐥 In plain terms: MCP is a server that sits between your AI agent and the outside world, acting as a bridge that lets your agent read information from tools and take actions inside them. Not just read. Actually #dothings . Let us take an example. You want your agent to draft emails and either save them or send them on your behalf. Here's what needs to happen: Your agent needs to 𝐭𝐚𝐥𝐤 to Gmail Gmail needs to 𝐮𝐧𝐝𝐞𝐫𝐬𝐭𝐚𝐧𝐝 what the agent wants The action needs to be 𝐞𝐱𝐞𝐜𝐮𝐭𝐞𝐝 inside Gmail These three things don't happen automatically. Gmail doesn't speak AI. Your agent doesn't speak Gmail. Someone needs to translate. 𝐓𝐡𝐚𝐭 𝐭𝐫𝐚𝐧𝐬𝐥𝐚𝐭𝐨𝐫 𝐢𝐬 𝐭𝐡𝐞 𝐌𝐂𝐏 𝐬𝐞𝐫𝐯𝐞𝐫. 𝐒𝐨 𝐰𝐡𝐚𝐭 𝐚𝐜𝐭𝐮𝐚𝐥𝐥𝐲 𝐡𝐚𝐩𝐩𝐞𝐧𝐬 𝐬𝐭𝐞𝐩 𝐛𝐲 𝐬𝐭𝐞𝐩? 1. You tell your agent: "Draft a follow-up email to the client and save it." 2. Agent decides it needs Gmail access 3. Sends request to the Gmail MCP server 4. MCP server translates that into something Gmail understands 5. Gmail executes and the draft gets saved 6. MCP server confirms back to the agent 7. Agent tells you: "Done. Draft saved." You never touched Gmail. The agent handled it end to end. 🎯 Doesn’t this sound similar to “API integration”, 2 systems talking with each other and sharing details? 𝐖𝐡𝐚𝐭 𝐢𝐟 𝐲𝐨𝐮 𝐝𝐨𝐧'𝐭 𝐮𝐬𝐞 𝐌𝐂𝐏 𝐚𝐧𝐝 𝐣𝐮𝐬𝐭 𝐮𝐬𝐞 𝐀𝐏𝐈 𝐤𝐞𝐲𝐬 𝐢𝐧𝐬𝐭𝐞𝐚𝐝? API keys work, but they come with a cost. Every tool needs its own unique key, its own authentication flow, its own custom code to handle responses and errors. Connect Gmail, Slack, Notion, and Calendar separately and you've built four different integrations, each one maintained independently. It's like instead of one universal power adapter, you're carrying four different country-specific plugs in your bag. Works. But painful. ⚡ #AILiteracy #AIPM #API #ProductManagement