Jiaqi Zhai
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Jiaqi Zhai has a total of 81 co-authors. Here are the top co-authors by number of shared publications:
- Xing Liu
Meta Recommendation Systems, Menlo Park, CA, USA
4 shared publications - Fu Li
Meta Platforms, Inc., Menlo Park, California, USA
2 shared publications - Yueming Wang
Meta Platforms, Inc., Menlo Park, California, USA
2 shared publications - Zhaojie Gong
Meta Platforms, Inc., Menlo Park, California, USA
2 shared publications - Alicia Golden
Meta, Cambridge, MA, USA
1 shared publication - Anna Sun
Meta, Menlo Park, CA, USA
1 shared publication
Latest Publications
- Open Access
- 19 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 4616-4621
Request-Only Optimization for Recommendation Systems
- 19 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 4616-4621
Open AccessRequest-Only Optimization for Recommendation Systems
- 8 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Micro
- Vol. 45 (5) , 103-115
Characterizing and Efficiently Accelerating Multimodal Generation Model Inference
Generative artificial intelligence (AI) technology is revolutionizing the computing industry, posing new system design and optimization opportunities. In particular, AI’s ability to understand and respond in multiple modalities comes with significant system resource demands. To sustainably scale generative AI capabilities to billions of users in the world, inference must be fast and efficient. This paper pinpoints key system design and optimization opportunities by characterizing a family of emerging multi-modal generation models on real systems. Auto-regressive token generation is a critical latency performance bottleneck, typically dominated by GPU idle time. In addition to memory-intensive attention across the generative AI models, linear operations constitute significant inference latency due to the feed forward networks in Transformer-based models. We demonstrate that state-of-the-art optimization levers, spanning from applications to system software and hardware, set a 3.88× better baseline.
- 8 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Micro
- Vol. 45 (5) , 103-115
Characterizing and Efficiently Accelerating Multimodal Generation Model Inference
Generative artificial intelligence (AI) technology is revolutionizing the computing industry, posing new system design and optimization opportunities. In particular, AI’s ability to understand and respond in multiple modalities comes with significant system resource demands. To sustainably scale generative AI capabilities to billions of users in the world, inference must be fast and efficient. This paper pinpoints key system design and optimization opportunities by characterizing a family of emerging multi-modal generation models on real systems. Auto-regressive token generation is a critical latency performance bottleneck, typically dominated by GPU idle time. In addition to memory-intensive attention across the generative AI models, linear operations constitute significant inference latency due to the feed forward networks in Transformer-based models. We demonstrate that state-of-the-art optimization levers, spanning from applications to system software and hardware, set a 3.88× better baseline.
- Open Access
- 22 April 2025
- Published by Association for Computing Machinery (ACM)
- p. 1626-1637
Retrieval with Learned Similarities
- 22 April 2025
- Published by Association for Computing Machinery (ACM)
- p. 1626-1637
Open AccessRetrieval with Learned Similarities
- Open Access
- 8 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 778-780
Enhancing Performance and Scalability of Large-Scale Recommendation Systems with Jagged Flash Attention
The integration of hardware accelerators has significantly advanced the capabilities of modern recommendation systems, enabling the exploration of complex ranking paradigms previously deemed impractical. However, the GPU-based computational costs present substantial challenges. In this paper, we demonstrate our development of an efficiency-driven approach to explore these paradigms, moving beyond traditional reliance on native PyTorch modules. We address the specific challenges posed by ranking models’ dependence on categorical features, which vary in length and complicate GPU utilization. We introduce Jagged Feature Interaction Kernels, a novel method designed to extract fine-grained insights from long categorical features through efficient handling of dynamically sized tensors. We further enhance the performance of attention mechanisms by integrating Jagged tensors with Flash Attention. Our novel Jagged Flash Attention achieves up to 9 × speedup and 22 × memory reduction compared to dense attention. Notably, it also outperforms dense flash attention, with up to 3 × speedup and 53% more memory efficiency. In production models, we observe 10% QPS improvement and 18% memory savings, enabling us to scale our recommendation systems with longer features and more complex architectures.
- 8 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 778-780
Open AccessEnhancing Performance and Scalability of Large-Scale Recommendation Systems with Jagged Flash Attention
The integration of hardware accelerators has significantly advanced the capabilities of modern recommendation systems, enabling the exploration of complex ranking paradigms previously deemed impractical. However, the GPU-based computational costs present substantial challenges. In this paper, we demonstrate our development of an efficiency-driven approach to explore these paradigms, moving beyond traditional reliance on native PyTorch modules. We address the specific challenges posed by ranking models’ dependence on categorical features, which vary in length and complicate GPU utilization. We introduce Jagged Feature Interaction Kernels, a novel method designed to extract fine-grained insights from long categorical features through efficient handling of dynamically sized tensors. We further enhance the performance of attention mechanisms by integrating Jagged tensors with Flash Attention. Our novel Jagged Flash Attention achieves up to 9 × speedup and 22 × memory reduction compared to dense attention. Notably, it also outperforms dense flash attention, with up to 3 × speedup and 53% more memory efficiency. In production models, we observe 10% QPS improvement and 18% memory savings, enabling us to scale our recommendation systems with longer features and more complex architectures.
- Open Access
- 4 August 2023
- Published by Association for Computing Machinery (ACM)
- p. 5520-5531
Revisiting Neural Retrieval on Accelerators
Retrieval finds a small number of relevant candidates from a large corpus for information retrieval and recommendation applications. A key component of retrieval is to model (user, item) similarity, which is commonly represented as the dot product of two learned embeddings. This formulation permits efficient inference, commonly known as Maximum Inner Product Search (MIPS). Despite its popularity, dot products cannot capture complex user-item interactions, which are multifaceted and likely high rank. We hence examine non-dot-product retrieval settings on accelerators, and propose mixture of logits (MoL), which models (user, item) similarity as an adaptive composition of elementary similarity functions. This new formulation is expressive, capable of modeling high rank (user, item) interactions, and further generalizes to the long tail. When combined with a hierarchical retrieval strategy, h-indexer, we are able to scale up MoL to 100M corpus on a single GPU with latency comparable to MIPS baselines. On public datasets, our approach leads to uplifts of up to 77.3% in hit rate (HR). Experiments on a large recommendation surface at Meta showed strong metric gains and reduced popularity bias, validating the proposed approach's performance and improved generalization.
- 4 August 2023
- Published by Association for Computing Machinery (ACM)
- p. 5520-5531
Open AccessRevisiting Neural Retrieval on Accelerators
Retrieval finds a small number of relevant candidates from a large corpus for information retrieval and recommendation applications. A key component of retrieval is to model (user, item) similarity, which is commonly represented as the dot product of two learned embeddings. This formulation permits efficient inference, commonly known as Maximum Inner Product Search (MIPS). Despite its popularity, dot products cannot capture complex user-item interactions, which are multifaceted and likely high rank. We hence examine non-dot-product retrieval settings on accelerators, and propose mixture of logits (MoL), which models (user, item) similarity as an adaptive composition of elementary similarity functions. This new formulation is expressive, capable of modeling high rank (user, item) interactions, and further generalizes to the long tail. When combined with a hierarchical retrieval strategy, h-indexer, we are able to scale up MoL to 100M corpus on a single GPU with latency comparable to MIPS baselines. On public datasets, our approach leads to uplifts of up to 77.3% in hit rate (HR). Experiments on a large recommendation surface at Meta showed strong metric gains and reduced popularity bias, validating the proposed approach's performance and improved generalization.
- Open Access
- 1 July 2022
- Published by Institution of Engineering and Technology (IET) in Chinese Journal of Electronics
- Vol. 31 (4) , 764-772
MSK‐PK: A Public‐Key Encryption Cryptosystem with Multiple Secret‐Keys
- 1 July 2022
- Published by Institution of Engineering and Technology (IET) in Chinese Journal of Electronics
- Vol. 31 (4) , 764-772
Open AccessMSK‐PK: A Public‐Key Encryption Cryptosystem with Multiple Secret‐Keys