InMobi

InMobi Advertising and Glance will be brewing interesting things at #WPPBrew tomorrow! Visit us at the experience zone, where we will showcase innovations across agentic shopping, mobile lock screen, the open internet, and connected TV. We’ve got cold brew and great conversations waiting. 🙌☕ Sakina Ghiya | Kamakshi Kamath | Priyanka Narasimhalu | Sairam Anchalkar

In the world of LLMs, memory is the problem hiding in plain sight. Everyone likes to focus on compute. But when LLMs move from demos to deployment, memory becomes the real bottleneck. High Bandwidth Memory (HBM) is now the costliest part of AI hardware, tightly linking players like NVIDIA and Samsung Semiconductor . KV cache grows rapidly, RAM comes under pressure, and the economics of serving models shift quickly. This is why Google 's breakthrough, TurboQuant, made waves in the tech ecosystem. The paper showed that the KV cache, the largest memory consumer in AI inference, can be compressed to just 3 bits, enabling up to a 6x memory reduction. Our very own InMobi engineer, Vibhav Agarwal , took this research from paper to production, wrote a full implementation, battle-tested it across model architectures and high-concurrency workloads, and got it merged into vLLM, the world's most popular open-source LLM framework. This took persistence, with 20–30 review cycles, close collaboration with maintainers from Red Hat , and participation in their weekly forums. They praised and approved his work for its clean, standalone design, making it easy to extend and future-proof without risking core systems. His architecture also crossed hardware lines, with teams at AMD and Intel already adapting it to their GPU environments, with more contributions underway. Because of his work, developers globally can reduce AI memory usage by up to 4.9x with a simple config switch, run 3–4x more requests on a single GPU than earlier required across four GPUs, and support longer context calls without running into OOM issues. At InMobi , we thrive on a fiercely entrepreneurial culture, built on an “In it together” philosophy that drives ownership and strong results. This is what it looks like in action. Our people make the extraordinary feel everyday. Take a bow, Vibhav. Never stop making us proud. Naveen Tewari Shefali Rai Arvind Jayaprakash Jairaj Sathyanarayana Chid Kollengode Ian Anderson Satyen Abrol Debleena Das Reneta K. Iliyas Shirol

Some of the amazing work happening at InMobi Glance.

A question I keep coming back to, and one that comes up in almost every conversation I have these days: Is AI an existential threat to the advertising industry? We've been here before. When machines arrived, the prediction was that humans would lose all their jobs. Didn't happen. When digital advertising emerged, the prediction was that growth would stall. Didn't happen. When mobile took over, the prediction was that TV advertising would collapse. Didn't happen. When social media exploded, the prediction was that attention spans would shrink too much to advertise to. Didn't happen. As human beings, we almost always overestimate short-term disruption and underestimate long-term transformation. And AI is no different. Here is what I think it actually unlocks for advertising: 1. More time means more content means more consumption As AI makes daily life more efficient, people will reclaim time and spend it on things they genuinely love. More creativity, more digital media, more content. And more content has always needed a business model to fund it. Advertising has always been that model, and it will continue to be. 2. Advertising will finally be held to the standard it always promised Digital brought measurement into the equation. AI takes it further. Advertisers will be able to prove, with far more precision, what is working and what is not. That is not a threat to advertising. That is what makes advertising worth investing in. 3. The future of advertising is agent discovery Every brand will have its own agent. Every consumer will have their own agent. The next frontier is facilitating those agent-to-agent interactions. Discovery, in a world of agents, becomes the new prime real estate. That is exactly what we are building at Glance . The opportunity ahead is bigger than anything the industry has seen before. That energy is what drives everything we do at InMobi .