Althaf Khan Pattan is a Senior Front-End Engineer and Tech Lead at Comcast, where he leads distributed engineering teams across the US and India, delivering large-scale consumer-facing web and mobile platforms. Alongside his industry role, he is an independent researcher working at the intersection of applied machine learning and front-end software architecture, with a focus on reinforcement learning, predictive rendering, and resilient system design for large-scale web applications. His research explores how ML techniques traditionally used in recommendation and search can be applied to UI-level decisions such as layout optimization and content preloading, bridging original research with the realities of production systems operating at scale.

Title: Reinforcement Learning for Full-Page Layout Optimization in E-Commerce Conversion Flows - Abstract:Static page layouts in e-commerce applications are typically chosen through A/B testing or hand-tuned heuristics, both of which struggle to adapt to the diversity of user contexts, device types, and shifting behavior patterns seen in production traffic. This talk presents a reinforcement learning approach to full-page layout optimization, where layout decisions are treated as a sequential decision-making problem rather than a fixed configuration choice. I will walk through how the layout selection problem is framed as a Markov Decision Process, how state representation captures session and user context, and how reward signals are designed around conversion-oriented outcomes such as click-through and add-to-cart behavior. The talk covers the tradeoffs between online and offline RL training in a production e-commerce setting, including the risks of live exploration against real users and strategies for safe policy evaluation before deployment. I will also discuss how this approach compares to traditional A/B testing and bandit-based methods, and where each remains preferable depending on traffic volume and business risk tolerance. Attendees will leave with a practical framework for evaluating whether RL-based layout optimization fits their own e-commerce architecture, and what infrastructure and data prerequisites are needed to get there safely.