Applied Machine Learning (AML) research
Expanding the Pareto frontier of AI. We translate theoretical breakthroughs into planetary-scale impact by making frontier intelligence computationally viable and seamlessly deployable.
About us
The future of AI cannot rely solely on scaling physical hardware; it requires algorithmic efficiency. We bridge the critical gap between fundamental AI discovery and global product realities. By pioneering breakthroughs in algorithmic efficiency, extreme model compression, and agentic systems, we decouple model intelligence from hardware constraints.
Small & efficient models
Designing highly optimized, sub-4B parameter architectures (including dense and MoE encoders/decoders) to deliver frontier-level intelligence with ultra-low latency and minimal memory footprints.
Generative media and multimodal understanding
Advancing the state-of-the-art in image, video, and audio generation through extreme step-reduction and knowledge distillation for models like Imagen, Lyria, Omni, and Gemini.
Parameter-efficient tuning & adaptation
Democratizing model adaptation via efficient tuning and reinforcement learning (e.g., LoRA, DPO), enabling rapid, cost-effective customization for enterprise and consumer applications.
Agentic systems & context optimization
Building autonomous agent frameworks, automated prompt optimization loops, and active memory management techniques to scale complex, long-horizon AI workflows.
Foundation models for personalization
Developing unified, cross-surface user models that capture deep behavioral intent, bootstrapping discovery and redefining the architecture of modern recommender systems.
Algorithmic efficiency & compression
Developing advanced model surgery techniques—including sequence compression, hardware-aware quantization, and knowledge decoupling—to drastically reduce memory constraints and compute waste.
“The future of AI requires algorithmic efficiency, not just hardware scaling. AML Research pushes the computational Pareto frontier, translating theoretical breakthroughs into the optimized algorithms and infrastructure powering Google’s core products.”
Featured research
Type-Compliant Adaptation Cascades
Chu-Cheng Lin, Daiyi Peng, Yifeng Lu, Ming Zhang, Eugene Ie
-
CVPR 2026
Designing Instance-Level Sampling Schedules via REINFORCE with James-Stein Shrinkage
Peiyu Yu, Suraj Kothawade, Sirui Xie, Ying Nian Wu, Hongliang Fei.
-
ICML 2026
Holistic Latent Diffusion Acceleration: Unifying Spatial, Temporal, and Architectural Efficiency
Ruyi An, Xin Yuan, Xixi Hu, Yeqing Li, Eugene Ie, Hongliang Fei, Mingyuan Zhou, Keyang Xu.
-
ICLR 2026
One Model, Many Goals: Meta-Learning Preference-Conditioned Alignment for Lifelong LLM Agents
Fatemeh Nourzad, Daouda Sow, Yingbin Liang, Ming Shi, Ming Zhang, Yunxuan Li, Eylem Ekici, Ness Shroff.
-
CVPR 2025
AMO Sampler: Enhancing Text Rendering with Overshooting
Xixi Hu, Keyang Xu, Bo Liu, Qiang Liu, Hongliang Fei.
-
ICCV 2025
Improving Rectified Flow with Boundary Conditions
Xixi Hu, Runlong Liao, Keyang Xu, Bo Liu, Yeqing Li, Eugene Ie, Hongliang Fei, Qiang Liu.
-
NeurIPS 2024
Subject-driven Text-to-Image Generation via Preference-based Reinforcement Learning.
Yanting Miao, William Loh, Suraj Kothawade, Pascal Poupart, Abdullah Rashwan, Yeqing Li.
-
ACM RecSys 2024
Calrec: Contrastive alignment of generative LLMs for sequential recommendation.
Yaoyiran Li, Xiang Zhai, Moustafa Alzantot, Keyi Yu, Ivan Vulić, Anna Korhonen, Mohamed Hammad.
-
arXiv 2026
How Fast Should a Model Commit to Supervision? Training Reasoning Models on the JQ Loss Continuum.
Chu-Cheng Lin, Eugene Ie
-
EMNLP 2024
Improving Multi-Agent Debate with Sparse Communication Topology
Yunxuan Li, Yibing Du, Jiageng Zhang, Le Hou, Peter Grabowski, Yeqing Li, Eugene Ie.
-
WACV 2024
MaskConver: Revisiting Pure Convolution Model for Panoptic Segmentation.
Abdullah Rashwan, Jiageng Zhang, Ali Taalimi, Fan Yang, Xingyi Zhou, Chaochao Yan, Liang-Chieh Chen, Yeqing Li.
Publications
Lin, C. C., & Ie, E., 2026
Learning Adaptive Reasoning Budgets via Constraint-Rectified Training.
Wu, Q., Lin, S., Zhang, M., Liang, Y., & Shroff, N. B., 2026
Constraint-rectified training for efficient chain-of-thought.
Wu, Q., Lin, S., Zhang, M., Liang, Y., & Shroff, N. B., 2026
On Surprising Effectiveness of Masking Updates in Adaptive Optimizers.
Joo, T., Xia, W., Kim, C., Zhang, M., & Ie, E., 2026
Type-Compliant Adaptation Cascades.
Lin, C. C., Peng, D., Lu, Y., Zhang, M., & Ie, E., 2026
SEVerA: Verified Synthesis of Self-Evolving Agents.
Banerjee, D., Xu, C., Ie, E., Zhang, M., Peng, D., Lin, C. C., & Singh, G., 2026
One Model, Many Goals: Meta-Learning Preference-Conditioned Alignment for Lifelong LLM Agents.
Nourzad, F., Sow, D., Liang, Y., Shi, M., Zhang, M., Li, Y., ... & Shroff, N., 2026
Barriers to Pareto Steerability in Preference-Conditioned LLM Alignment.
Nourzad, F., Sow, D., Liang, Y., Shi, M., Zhang, M., Li, Y., ... & Shroff, N., 2026
Haoqin Tu, Jianwen Chen, Zijun Wang, Juncheng Wu, Hardy Chen, Haonian Ji, Kaiwen Xiong, Siwei Han, Jiaqi Liu, Peng Xia, Jieru Mei, Hongliang Fei, Jason Eshraghian, Zeyu Zheng, Yuyin Zhou, Huaxiu Yao, Cihang Xie, 2026
Holistic Latent Diffusion Acceleration: Unifying Spatial, Temporal, and Architectural Efficiency.
Ruyi An, Xin Yuan, Xixi Hu, Yeqing Li, Eugene Ie, Hongliang Fei, Mingyuan Zhou, Keyang Xu, 2026
Designing Instance-Level Sampling Schedules via REINFORCE with James-Stein Shrinkage.
Peiyu Yu, Suraj Kothawade, Sirui Xie, Ying Nian Wu, Hongliang Fei, 2026
Bridging Textual Profiles and Latent User Embeddings for Personalization.
Tan, Zhaoxuan, Xiang Zhai, Yan Zhu, Meng Jiang, and Mohamed Hammad, 2026
Improving Rectified Flow with Boundary Conditions.
Xixi Hu, Runlong Liao, Keyang Xu, Bo Liu, Yeqing Li, Eugene Ie, Hongliang Fei, Qiang Liu, 2025
AMO Sampler: Enhancing Text Rendering with Overshooting.
Xixi Hu, Keyang Xu, Bo Liu, Qiang Liu, Hongliang Fei, 2025
Improving Multi-Agent Debate with Sparse Communication Topology.
Yunxuan Li, Yibing Du, Jiageng Zhang, Le Hou, Peter Grabowski, Yeqing Li, Eugene Ie, 2024
Efficient Transformer Adaptation with Soft Token Merging.
Xin Yuan, Jinoo Baek, Hongliang Fei, 2024 (Pending conference submission / CVPR Workshops 2024)
Inducing Generalization across Languages and Tasks using Featurized Low-Rank Mixtures.
Lin, C. C., Wang, X., Clark, J. H., Lu, H., Zhu, Y., Whitehouse, C., & Yu, H., 2024
Subject-driven Text-to-Image Generation via Preference-based Reinforcement Learning.
Yanting Miao, William Loh, Suraj Kothawade, Pascal Poupart, Abdullah Rashwan, Yeqing Li, 2024
Greedy Growing Enables High-Resolution Pixel-Based Diffusion Models.
Cristina Nader Vasconcelos, Abdullah Rashwan, Austin Waters, Trevor Walker, Keyang Xu, Jimmy Yan, Rui Qian, Yeqing Li, Shixin Luo, Yasumasa Onoe, Zarana Parekh, Ivana Kajic, Mandy Guo, Wenlei Zhou, Sarah Rosston, Roopal Garg, Hongliang Fei, Jordi Pont-Tuset, Su Wang, Henna Nandwani, Andrew Bunner, Kevin Swersky, David J. Fleet, Oliver Wang, Jason M. Baldridge, 2024
Inflation With Diffusion: Efficient Temporal Adaptation for Text-to-Video Super-Resolution.
Xin Yuan, Jinoo Baek, Keyang Xu, Omer Tov, Hongliang Fei, 2024
MaskConver: Revisiting Pure Convolution Model for Panoptic Segmentation.
Abdullah Rashwan, Jiageng Zhang, Ali Taalimi, Fan Yang, Xingyi Zhou, Chaochao Yan, Liang-Chieh Chen, Yeqing Li, 2024
Calrec: Contrastive alignment of generative llms for sequential recommendation.
Li, Yaoyiran, Xiang Zhai, Moustafa Alzantot, Keyi Yu, Ivan Vulić, Anna Korhonen, and Mohamed Hammad, 2024
Embedding User-Generated Content using Structural Supervision and Generative Models.
Vinay Shukla, Yang Yang, Siddarth Malreddy, Jinoo Baek, Dale Johnson, Wenfei Zhou, Minh Pham, Mark Williams, Karthik Lakashman, 2023
Locations
-
Mountain View
-
San Francisco
-
Kirkland
-
New York
-
London