Tata Digital

‘IF YOU’RE NOT VIBE CODING, YOU ARE WAITING TO BE FIRED’ reads the billboard across design teams in India these days. When we are scared, we usually make wrong decisions. And if everyone around us is making similar decisions, it's just easy to go with the flow. That’s what is currently happening to designers fearful of AI. While the dust is still settling on where AI is taking us, everyone is looking at designers owning code as the ‘right decision’. Beer table conversations of designers sound like, ‘the dev team approved my code in one go’ or ‘a 0 bug interface is what I'm aiming for'. On the other side of the table, developers are blushing over the latest designs they’ve AI-created, the chic font that was chosen & discussing if they should update their LinkedIn profile, Full Stack *%^# Everything. The most fun is at the PM adda; they are jumping on hot tin roofs, designing, coding, and compulsively using AI even where it’s not needed. Hell Yeah! True democracy, finally, everybody is fighting for everybody’s seat. There is some divine justice in all of this, for far too long we’ve been working in silos, & the breaking of it is a welcome change. That’s how far it should go: break the silos, not yourself or the reason you chose your profession. When our current skills become easier to replicate, we need to go deeper, not always farther. There was a reason we chose this profession; let’s be true to that. Fellow Designers! We have a once-in-a-decade chance to relook into what will make us better designers & not just ‘AI Proof Designers’. Some areas within design, where we can now go deeper: Neuroscience: Dive into the science of hormones, not just the dopamine addiction, go deeper. 👤 Paul J. Zak Behavioural Psychology: Go beyond the simplistic understanding of System 1 and System 2. Three Nobel Prizes have been awarded in the last 15 years to Behavioural Economists; a lot is waiting to be put into design practice. 👤 Rory Sutherland , Dan Ariely , Stephen Wendel Business Storytelling: Don’t present your insights, own the room. 👤 David JP Phillips , Patsy Rodenburg https://lnkd.in/gMmv5Vss Facilitation: Move beyond brainstorming champions within the design team, be the organisation’s chosen workshop facilitator. 👤 Dave Gray https://gamestorming.com/ Accessibility: Be the accessibility champions, it's not just an add-on, it's a specialisation. 👤 Kat Holmes Speculation: Sense-making skills in this chaotic world will be at a premium. 👤 J. Paul Neeley Moral Imagination: Understanding the "why" behind long-term, empathy-driven design Phoebe Tickell Yes, I hear you, you find vibe coding liberating & love the experience of controlling the experience. Sure, keep at it & become better. More power to you. I am not against vibe coding & certainly for using AI tools for our work; I just feel that this is not what will safeguard our expertise. There’s enough within design to keep us not just relevant, but in demand.

Sea is quietly becoming one of the most important internet companies outside the US and China. And the fascinating part is this: Shopee may be the face of the business — but gaming and fintech are increasingly the engines keeping the entire machine running. This quarter made that very clear. Sea posted: • $7.1B in revenue (+46.6% YoY) • $3.1B gross profit • $438M net income • $1B adjusted EBITDA Those are monster numbers by any standard. But the real story sits beneath them. Shopee continues to dominate Southeast Asian ecommerce, yet growth is getting more expensive. Logistics costs are rising. Competition is intensifying. Margins are compressing. Gross margin fell from 46.2% to 44.3%. Ordinarily, markets would punish that. But Sea has something most ecommerce companies don’t: a cash-generating gaming business and a rapidly scaling fintech arm. Garena — the business many had written off after its pandemic peak — just posted its best quarter since 2021. Free Fire and Arena of Valor continue throwing off cash nearly a decade after launch. That matters because it gives Sea room to keep investing aggressively into Shopee while competitors burn capital externally. Meanwhile, Monee is quietly becoming a giant. Its loanbook crossed $1B in Brazil, making it the fourth Sea market to hit that scale. And loan growth there surged 250% YoY. This is where the Sea model becomes interesting. Ecommerce attracts users. Gaming generates cash. Fintech monetizes engagement. Together, they create a flywheel few companies globally have managed to build successfully. There’s another subtle point here too. Southeast Asia’s ecommerce market is maturing. TikTok Shop is investing aggressively. Alibaba continues backing Lazada. The region is no longer an easy growth story. Which is why Brazil matters so much. Shopee entered Brazil years ago when most people saw it as a side experiment. Today, it’s one of the company’s fastest-growing markets — and importantly, still profitable despite heavy logistics investment. Sea is no longer just defending Southeast Asia. It’s exporting its operating model globally. And unlike many internet companies chasing growth at all costs, Sea is proving something increasingly rare in tech: You can still grow fast, defend market share, invest aggressively — and remain profitable — if your ecosystem is deep enough.

I’ve started tipping "before" booking Uber rides now. Not after the ride. Before starting it. A friend told me: “Add a tip before requesting and you’ll get one faster.” I tried it, it worked, and it felt slightly odd. Then I came across this comment from Dr.Rama Moondra on my last post about UPI and “default generosity”: “...And that generosity for Uber , Ola and Rapido is compulsion. They won't accept the ride until you are topping it up.” I don’t know if that’s universally true, but I had noticed it too. Rides were taking longer to get accepted. The tip feature was actually smart product design. When demand spikes, riders could pay a little extra to get a quicker ride, while drivers earned a bit more too. But over time, drivers figured out the other side of that equation. During peak hours, tipped rides naturally become more attractive. Waiting becomes the rational choice. Hard to really blame anyone for that. The platform built a nudge and the market rewrote the rules around it. What started as a reward has quietly become an entry fee. Funny how quickly an optional feature can stop feeling optional. Do you tip before booking now, or still wait till the ride ends?

Hospitality may become the industry where AI delivers the most personalized experience at scale. Because unlike most industries, hospitality sits at the intersection of: behaviour, emotions, preferences, timing and real-world experiences. A few years back, hotel tech products were mostly solving surface-level problems: bookings, billing, channel management. But the real challenge inside hospitality was always operational chaos and inconsistent guest experience. And now AI is slowly moving into that layer. One example I recently came across was Treebo’s AI-powered hotel operating system “Hotel Superhero” originally built to manage operations across 750+ hotels. What stood out wasn’t just automation. It was the possibility of personalization and operational intelligence working together: - understanding repeat guest preferences, - predicting customer needs, - reducing service inconsistency, - simplifying front-desk operations, - helping owners monitor outlet performance centrally, - and reducing dependency on fragmented communication. That’s where AI becomes truly powerful in hospitality. Because a guest rarely remembers only the room or food. They remember: - how quickly issues were resolved, - whether preferences were remembered, - how personalized the experience felt, - and how consistently the service was delivered. Most hospitality businesses don’t fail because of lack of effort. They struggle because operations and personalization become difficult to scale together. One missed escalation. One wrong order. One delayed response. One poor handoff between teams. Over time, these small inefficiencies quietly impact customer trust, ratings and repeat business. I feel the next big advantage in hospitality may not come only from better interiors or larger marketing budgets. It may come from: AI-driven personalization + operational intelligence. Not replacing humans. But helping businesses deliver consistency and personalization at scale. #AI #Hospitality #Hotels #Restaurants #CustomerExperience #Personalization #Automation #Operations #FoodBusiness #Startups

Excited to see Super.money and our journey being featured on CNBC-TV18 Storyboard18. The conversation captures how fintech is evolving beyond just cashback and CAC-led growth, towards building scalable credit, payments and consumer engagement ecosystems for the next generation of users. A lot of exciting work happening across lending, UPI, partnerships, distribution and customer experience — with the vision of making credit more accessible, contextual and seamless.

I joined InMobi group on the Glance side a few weeks back. People, tech and the problem statements are what attracted me (plus I get to work with Shubha again:)). I can see that the talent density, pace and energy are phenomenal. I’ve joined the applied science / research team and I’m humbled by the amount and pace of work that has happened/ happening. Our work spans both hardware (GPUs) and software across several areas of ML/DL/CV. Led by Ian Anderson and Satyen Abrol , here is a sample of recent/ upcoming work by the team. Lots of work at the intersection of user modeling, computer vision, language models, recsys, clever engineering and (I guess:)) AI. > 'LookSync: Large-Scale Visual Product Search System for AI-Generated Fashion Looks' presented at International Conference on Data Science (IKDD CODS 2025). Nov 2025. > 'Synthetic Data for Virtual Try-On: Methodology and Lessons Learned' presented at SynIRgy Simulation and Synthetic Data for Information Retrieval (worksop part of European Conference on Information Retrieval). Apr 2026. > 'Understanding User Responses to Virtual Try-On Through Large-Scale and Qualitative Studies' presented at Fashion-Textiles-Wearables (FTW). Apr 2026. > 'Short-Lived High-Volume Bandits' published in INFORMS Operations Research. Apr 2026. > 'Cross-Domain Cold-Start Personalization via LLM-Synthesized Structured User Profiles' accepted at ACM UMAP 2026. Jun 2026. > 'Socio-Technical Insights from Virtual Try-On: Representation, Bias, and Trust' accepted at the 8th Asia Conference on Machine Learning and Computing. Jul 2026. > 'Identity-Aware Image Fidelity: A Perceptual Framework for Evaluating Personalised Generative Media' accepted at the 8th Asia Conference on Machine Learning and Computing. Jul 2026. > An open-source contribution implementing TurboQuant into vLLM. Apr 2026. Not to mention a few more publications under submission. I am already knee-deep in several initiatives. We are hiring strong applied/ research/ data scientists, ML engineers and full-stack builders with deep product and systems thinking, across levels. Please reach out. Mohit Saxena Arvind Jayaprakash Mansi Jain Shefali Rai Debleena Das .

₹𝟮𝟮,𝟵𝟯𝟭 𝗰𝗿𝗼𝗿𝗲 𝗹𝗼𝘀𝘁 𝘁𝗼 𝗱𝗶𝗴𝗶𝘁𝗮𝗹 𝗳𝗿𝗮𝘂𝗱 𝗶𝗻 𝟮𝟬𝟮𝟱. 𝗠𝗼𝘀𝘁 𝗼𝗳 𝘂𝘀 𝗸𝗻𝗼𝘄 𝘀𝗼𝗺𝗲𝗼𝗻𝗲 𝗶𝗻 𝘁𝗵𝗮𝘁 𝗻𝘂𝗺𝗯𝗲𝗿. Four years ago, it was a fraction of this. The scale of the jump is what should alarm us. The RBI has noticed and recently released a discussion paper proposing a few safeguards worth knowing about: One hour lag on transfers above ₹10,000, giving you time to cancel if something feels wrong. A kill switch to instantly disable all digital payments from your account. A trusted person to approve high value transfers for citizens above 70 and people with disabilities. A credit limit on accounts, beyond which funds are held until verified. None of these are final yet. But the direction is clear. We spent a decade making payments frictionless. Unfortunately, The fraudsters also benefited due to this. The next decade will be about finding the right equilibrium. Fast enough to be useful. Slow enough to be safe. Where should the system tilt? Speed, even if some fraud is inevitable? Or safety, even if it breaks the magic of instant payments?

I asked AI to write my LinkedIn post AI asked me to review it I asked AI to improve my review AI asked me to approve the improvement I asked AI if the approval sounded right AI suggested a few tweaks I asked AI to review the tweaks It has been 3 hours and I have completely lost track of who is working for whom. The post is still not out and hence, I had to post this instead. #GenAI #AIHumour #FutureOfWork #Productivity #AI

Aaron Levie 's observation about software going headless warrants more attention. The agent era requires enterprises to re-evaluate years of architectural choices: Enterprise software moats were built on human adoption for the longest time - certified admins, trained users, and usage habits. Agents don't have those. Switching costs were real because humans bore them in relearning, retraining and rebuilding muscle memory (and the parallel economy of trainings!). Agents don't need that kind of cost. The evaluation criteria shift from "which platform do your people prefer?" to "which data model do your agents perform best on?" The complexity of using the software was a human moat that has now become a liability when agents navigate it. I have been following David Heinemeier Hansson at 37signals for years, and they have been making similar architectural bets for two decades, including their latest push to make their suite API-led for agents. Their decisions may have sounded off at the time, but they look well-suited to the new world. Also, for the enterprises consuming these APIs, there is a new set of questions to address now: 1. The vendor contracts were priced for human-scale usage. Agent-scale consumption is a grey zone 2. They could train humans and hold them accountable. Agents need guardrails and architectural governance, and a misconfigured agent doesn't make one mistake; it makes the same mistake 100,000 times. 3. Regulators expect human accountability chains. Agents break those chains. The architectural opportunity is real for both vendors and enterprises. The governance gap is equally real.