Suchhanda Das

❓ "So how is it going on the new job?" ⁉️ Since that is a question I'm getting a lot nowadays why not make it public! I'm already in my second month at CrazyGames and it's been simply amazing so far: * There are so many talented people, each with their own unique skillset which makes every interaction you have interesting * There is a lot of trust and ownership with low stress levels and a lot of focus on having fun and keeping healthy * We have a remote first policy but everyone is making active efforts to stay connected and is all-in on our mission to become the worlds biggest casual gaming platform In order to achieve that mission we are still growing, adding new expertises to the team. Next week we'll have 10 new people onboarded since I joined 🤯. Next to recruitment I also have the chance to look at other areas of talent management within the company which makes it even more varied. I must say switching from the IT consultancy sector to the gaming industry has been an unexpected but welcome change. Looking at how our product takes shape and seeing the daily efforts being handled with a lot of fun (sometimes I honestly do not know if people are working or just playing some games 😁) is a fantastic experience. Instead of chasing quarterly deal targets, I get the chance to help build something for the long run and that sense of purpose is intoxicating. Keep an eye on this space as I suspect we will be hiring for some cool positions soon 👀! I added some pictures of our boulder session yesterday with the international and newly formed advertising team which came over to Belgium last week to get acquainted. #CrazyGames #GamingJobs

CrazyGames broke into the top 400 most visited websites globally. We're now the 386th most visited website in the world, just ahead of Nintendo.com and before Xbox.com. We'll gladly take our place between two gaming giants. 💜 And in case you're wondering, Crazy Ziggy hasn't let go of that football since the start of the World Cup. ⚽

CrazyGames broke into the top 400 of most visited websites globally. We're now the 386th most visited website in the world, right after Xbox.com and just ahead of Nintendo.com 😍

A layoff. A restart. Another leap into the unknown. And through it all — the decision to keep building. Episode 4 of Beyond the Build powered by SeedlingLabs is now live. In this episode, our CTO, Sumanta Ranjan Das, joins our Founder & CEO, Shanti Kuropati, for a deeply honest conversation on resilience, reinvention, and navigating uncertainty across nearly three decades in technology. From growing up in Ranchi and studying at IIT Kharagpur, to experiencing layoffs during defining moments in India’s tech evolution, Sumanta reflects on the experiences that shaped his philosophy toward leadership, curiosity, and continuous adaptation. “I don’t just believe in exploring the unknown. I believe in embracing the unknown.” This is a conversation about staying grounded while the world changes around you. About continuing to evolve through every technological wave. And about building forward — even when the future feels uncertain. Watch the full episode: https://lnkd.in/gGsBmfee

Last week I won Nirvana Insurance's Pinnacle Award, one of six peer-nominated awards given across the company each year. In late February I started work on Nirvana AI, our internal AI platform that lets anyone at the company pull data, build tools, and automate workflows without writing code or filing an engineering ticket. Under the hood, the big technical bet was letting the agent write Python against a custom SDK instead of relying on structured tool calls. We stripped MCP from the agent's primary path and replaced it with a 14,000-line Python SDK that wraps every system the platform touches. Code execution is the only way the agent moves data, builds charts, or ships apps. Anthropic's November post on code execution with MCP (https://lnkd.in/gsF-Pfbk) showed a 98.7% token reduction by having the model write code instead of calling tools directly. Cloudflare independently landed on the same idea and called it "Code Mode" (https://lnkd.in/g9fMXiJ7). In late December, a paper on recursive language models (https://lnkd.in/gmRECwB3) went further: an 8B model that programmatically decomposed prompts instead of stuffing them into context outperformed its own baseline by 28% and approached GPT-5 quality on three long-context tasks. We shipped the beta, and then I spent eight weeks on the road working with teams in SF, New York, Columbus and Atlanta. The best part was watching people who had never written code build real things fast. Compliance has form accuracy monitors and filing dashboards running daily. Underwriting built priority queues, hourly authority alerts, and driver review agents firing every two hours. Billing has ops agents handling reconciliation and payment reminders. BD ships agency QBR generators. Pricing has precision pricing explorers. Every department now has someone shipping their own tooling. A lot of why this works is Nirvana itself. The company gives every employee, not just engineers, direct access to AI tools and lets them ship. And people across every team have been unusually tech-forward, sending bug reports at 11pm and feature requests in real time, with a constant stream of feedback that has made the platform sharper every week. Someone notices something worth automating at 3pm and has a working version live by 4pm. We're constantly breaking, fixing, and learning under real load. But the traffic tells the story. First 30 days, Feb 24 to Mar 24: 9,871 conversations across 171 users, 52,053 messages, 6,576 artifacts, 127,622 agent code executions, 359 reusable workflows, 91 scheduled automations, and 240 live Nirvana Apps. We've since crossed 170,000 messages, 850 apps, and 200+ scheduled automations, and growing. More to share soon on the architecture, what we've broken, and where this is going. Also, shoutout to Puneet Bhatt and Shashank M., who I'm lucky to keep learning from and trading AI and Engineering ideas with outside of work.

This week, I attended my first Nirvana Insurance offsite in Bangkok since joining in September. Over two days, the focus was less on presentations and more on bringing teams together around where R&D is headed in 2026, and how engineering must evolve in an AI-native world. 🚀 I also had the opportunity to host a panel discussion with Olivia (Beins) Cartwright, Rishi Swami, and Abhay Kumar - leaders across Product, Data Science, and Engineering. We explored some of the important questions many organizations are actively navigating today: • How do we ship faster without compromising quality? • When does a prototype become a product? • How do small, autonomous teams stay aligned without slowing down? • As AI accelerates output, what human skills become even more valuable? What stood out most was the shared focus on fundamentals: clear thinking, strong judgment, collaboration, deep context, and making sound decisions under ambiguity. AI will continue to reshape how we build but these fundamentals will determine how successfully we build. 🤝 Grateful for the thoughtful discussions, team connections, and the opportunity to learn from exceptional people across the organization. Excited for what lies ahead. ✨

P99s are like ghosts—sometimes you can’t explain them, but you can still defend against them. We’ve all been there: digging through logs to find out why one specific request took 2 seconds while the other 99 took 20ms. Sometimes it’s a GC pause, a noisy neighbor, or a temporary network blip. You can spend weeks debugging it, or you can use Request Hedging. Check it out: https://lnkd.in/g-XvXGUN #SystemDesign #Backend #Latency #SoftwareEngineering

92% faster updates. No more "Merge Wait." ⚡ As we grew, a "traffic jam" of code updates was slowing us down. The more we built, the longer the PR merge wait time became. So, we built a faster lane. We created NirvanaMQ, a custom system that slashed our merge wait times by 92%. Now, our engineers spend less time waiting and more time building. The best part? It started as a 10-week project and now handles thousands of updates a year. 🚀 👉 See how we cleared the path: https://lnkd.in/dg8epYw2