Serving deep learning based recommendation models (DLRM) at scale is challenging. Existing approaches rely on dedicated ANN indexing and filtering services on CPUs, suffering from non-negligible costs and missing co-design opportunities. Such inefficiency makes them difficult to support complex model architectures, such as learned similarities and multi-task retrieval. In this paper, we present SilverTorch, a model-based serving system that brings all components into one unified model. It unifies model serving by replacing standalone indexing and filtering services with model layers. We propose a model-based GPU Bloom index for feature filtering and a fused Int8 ANN kernel for nearest neighbor search. Through co-design of the ANN search and feature filtering, we reduce GPU memory usage and eliminate computation. Benefiting from this design, we scale up retrieval by introducing an OverArch scoring layer and a multi-task retrieval with a Value Model to aggregate scores. These advancements improve the retrieval accuracy and enable future studies for serving more complex models. Our evaluation on industry-scale datasets shows that SilverTorch achieves up to 23.7× higher throughput compared to the state-of-the-art approaches. We also demonstrate that SilverTorch's solution is 13.35× more cost-efficient than CPU-based solution while improving accuracy via serving more complex models.

This Research Work in Progress Paper presents a novel mobile software application created to perform ecological momentary assessment in two longitudinal studies with undergraduate engineering students. The studies collect learning activity data from first and second-year mechanical engineering students to examine the correlation of active learning with grit growth and student success and retention, and the possible associations between these factors. The software system consists of native Android and iOS mobile applications and a REST-based server application for prompting and collecting learning activity data from student participants. The system utilizes personalized student schedule data to push notifications to students on their smartphones as they complete each class or lab section. Students then submit the amount of time they spent in each class doing different types of activities including interactive, constructive, active, and passive. Students also receive notification reminders to enter out-of-class study activities. The system also includes automated daily audits and notifications to ensure consistent data entry activity. This paper will discuss the research context, early study results, software architecture, and the software development experience including institutional challenges, rationale for technology choices, strategies for developing research software with student developers, and lessons learned.