Aditya Mantha

Hanyu Li, Yi-Ping Hsu, Aditya Mantha et al. · stanford

Modern recommendation systems predominantly train retrieval and ranking as separate models despite both increasingly relying on large transformers encoding the same user behavior data, duplicating parameters, compute, and serving cost. Prior work unifies the model architecture but not the full pipeline: input formats, training procedures, and serving stacks remain fragmented across stages. We present UniPinRec, which achieves full-stack unification of retrieval and ranking at Pinterest: one input format, one model, one training stage, deployed within existing serving infrastructure. A shared transformer encodes the user action sequence into candidate-independent representations that branch into retrieval (ANN dot-product) and ranking (cross-attention) via task-specific heads. Three ideas make this work: (1) Masked Action Modeling (MAM) eliminates interleaving, enabling weight sharing without doubling context length; (2) Blended training examples pair action sequences with feedview impression slates to satisfy both objectives jointly; (3) Cross-stage KV cache sharing reuses user-history computation from retrieval for ranking, reducing total FLOPs versus serving two independent models. Deployed in the Pinterest core surfaces, UniPinRec delivers approximately +1% online engagement lift while cutting end-to-end serving latency by 11.1% and lifting QPS by 63.6%. To our knowledge, this is the first full-stack unification of retrieval and ranking, covering inputs, model, training and serving, deployed in a production recommendation system.

Shashank Kedia, Aditya Mantha, Sneha Gupta et al.

Through recent advancements in speech technologies and introduction of smart assistants, such as Amazon Alexa, Apple Siri and Google Home, increasing number of users are interacting with various applications through voice commands. E-commerce companies typically display short product titles on their webpages, either human-curated or algorithmically generated, when brevity is required. However, these titles are dissimilar from natural spoken language. For example, "Lucky Charms Gluten Free Break-fast Cereal, 20.5 oz a box Lucky Charms Gluten Free" is acceptable to display on a webpage, while a similar title cannot be used in a voice based text-to-speech application. In such conversational systems, an easy to comprehend sentence, such as "a 20.5 ounce box of lucky charms gluten free cereal" is preferred. Compared to display devices, where images and detailed product information can be presented to users, short titles for products which convey the most important information, are necessary when interfacing with voice assistants. We propose eBERT, a sequence-to-sequence approach by further pre-training the BERT embeddings on an e-commerce product description corpus, and then fine-tuning the resulting model to generate short, natural, spoken language titles from input web titles. Our extensive experiments on a real-world industry dataset, as well as human evaluation of model output, demonstrate that eBERT summarization outperforms comparable baseline models. Owing to the efficacy of the model, a version of this model has been deployed in real-world setting.

Aditya Mantha, Anirudha Sundaresan, Shashank Kedia et al.

E-commerce platforms consistently aim to provide personalized recommendations to drive user engagement, enhance overall user experience, and improve business metrics. Most e-commerce platforms contain multiple carousels on their homepage, each attempting to capture different facets of the shopping experience. Given varied user preferences, optimizing the placement of these carousels is critical for improved user satisfaction. Furthermore, items within a carousel may change dynamically based on sequential user actions, thus necessitating online ranking of carousels. In this work, we present a scalable end-to-end production system to optimally rank item-carousels in real-time on the Walmart online grocery homepage. The proposed system utilizes a novel model that captures the user's affinity for different carousels and their likelihood to interact with previously unseen items. Our system is flexible in design and is easily extendable to settings where page components need to be ranked. We provide the system architecture consisting of a model development phase and an online inference framework. To ensure low-latency, various optimizations across these stages are implemented. We conducted extensive online evaluations to benchmark against the prior experience. In production, our system resulted in an improvement in item discovery, an increase in online engagement, and a significant lift on add-to-carts (ATCs) per visitor on the homepage.

Mansi Ranjit Mane, Shashank Kedia, Aditya Mantha et al.

Through recent advancements in speech technology and introduction of smart devices, such as Amazon Alexa and Google Home, increasing number of users are interacting with applications through voice. E-commerce companies typically display short product titles on their webpages, either human-curated or algorithmically generated, when brevity is required, but these titles are dissimilar from natural spoken language. For example, "Lucky Charms Gluten Free Break-fast Cereal, 20.5 oz a box Lucky Charms Gluten Free" is acceptable to display on a webpage, but "a 20.5 ounce box of lucky charms gluten free cereal" is easier to comprehend over a conversational system. As compared to display devices, where images and detailed product information can be presented to users, short titles for products are necessary when interfacing with voice assistants. We propose a sequence-to-sequence approach using BERT to generate short, natural, spoken language titles from input web titles. Our extensive experiments on a real-world industry dataset and human evaluation of model outputs, demonstrate that BERT summarization outperforms comparable baseline models.

Aditya Mantha, Yokila Arora, Shubham Gupta et al.

With growing consumer adoption of online grocery shopping through platforms such as Amazon Fresh, Instacart, and Walmart Grocery, there is a pressing business need to provide relevant recommendations throughout the customer journey. In this paper, we introduce a production within-basket grocery recommendation system, RTT2Vec, which generates real-time personalized product recommendations to supplement the user's current grocery basket. We conduct extensive offline evaluation of our system and demonstrate a 9.4% uplift in prediction metrics over baseline state-of-the-art within-basket recommendation models. We also propose an approximate inference technique 11.6x times faster than exact inference approaches. In production, our system has resulted in an increase in average basket size, improved product discovery, and enabled faster user check-out