Beyond the model: Building real-world computer vision at Grab

Computer vision often looks neat in the lab, but deploying it at scale in the real world is a very different challenge. At Grab, we have spent the last decade building vision systems that operate under real constraints: limited compute, privacy requirements, diverse environments across Southeast Asia, and the need to turn perception into business value. This talk will trace that journey through Grab’s evolution in computer vision, powering map-making to OCR and multimodal content moderation for platform safety.

Al is only as good as the data behind it. At Grab, that means going beyond model building to engineer the full capture pipeline: our own camera systems, combined with multi-modal collection to cover different environments; and disciplined processes to make the data consistent, scalable, and production-ready. Successful computer vision is not just about better models, but also about building the data, systems, and operating discipline that make those models work in the real world.

Speaker’s profile

Sriram Iyer is a senior product and technology leader at Grab, an engineer­-turned product executive with 22+ years building and scaling real-world platforms. He has led globally distributed teams across the US, China, India, Singapore, and Romania, scaling from an IC to leading 130- to 150-person organisations, and has driven major 0➔1 and transformation initiatives. In this talk, he draws on that experience to explain how Computer Vision moves from models to measurable impact in messy, real-world environments – data, deployment, reliability, and the product decisions that make it stick.