Software Engineer @ Google · AI Researcher (Ph.D.)
Samiul Alam
Software Engineer at Google · AI & Spatial Computing Researcher (Ph.D., OSU)
I am a Software Engineer on the Shopping Personalization team at Google in Mountain View, CA, working on user activity modeling and agentic recommendation systems. Earlier, I interned with the same team working on synthetic user data generation.
I earned my Ph.D. in Computer Science and Engineering at The Ohio State University (2026) as a Graduate School University Fellow in the AIoT-MLSys Lab advised by Prof. Mi Zhang, collaborating with the Surreal team at Meta Reality Labs on agentic supermemory. In Summer 2026, I also interned as a Research Scientist with the Sensors and Systems Research team at Meta Reality Labs, working on on-sensor AI.
Previously, I earned my M.S. in CS at Michigan State University (2023) researching privacy-preserving distributed systems and federated learning, engineered wearable and 5G smart-router software at Samsung R&D (2018–2021), and co-founded Bengali.AI.
GoogleCurrentMeta Reality LabsThe Ohio State UniversityBengali.AI
Collaborated with the Surreal team at Meta Reality Labs during my Ph.D. to build agentic supermemory systems for AR smart glasses—spanning selective capture, gaze-guided retrieval, and long-horizon egocentric reasoning.
Agentic SupermemoryAR Smart GlassesEgocentric AI
Agentic & Multimodal Recommenders
Building user activity modeling and agentic recommendation systems on Google's Shopping Personalization team, alongside synthetic user data generation for personalization.
Agentic SystemsSynthetic User DataPersonalization
Efficient & On-Sensor AI
Designing resource-efficient foundation models and wearable intelligence—from singular-value truncation for LLM compression to on-sensor AI with Meta Reality Labs' Sensors and Systems Research team.
On-Sensor AIModel CompressionEdge Inference
Privacy-Preserving Distributed Systems
Designing model-heterogeneous, semi-asynchronous, and geo-distributed federated learning frameworks and authentic IoT benchmarks—enabling collaborative training across edge devices without centralizing private user data.
Joined the Shopping Personalization team at Google in Mountain View, CA full-time as a Software Engineer working on user activity modeling and agentic recommendation systems!
My Ph.D. dissertation was officially approved by The Ohio State University Graduate School and archived on OhioLINK!
Invited to serve on the Technical Program Committee (TPC) for IEEE MSN 2026 (22nd IEEE International Conference on Mobility, Sensing and Networking), alongside PC/reviewing service for ICLR 2027, AAAI 2027, and CVPR 2027.
Successfully defended my Ph.D. dissertation, “Towards Supermemory in Spatial Computing: A Multimodal Approach for Memory Creation, Retrieval, and Visual Reasoning,” at The Ohio State University (advised by Prof. Mi Zhang)!
Joined Meta Reality Labs in Redmond, WA as a Research Scientist Intern with the Sensors & Systems Research team, working on RL-driven egocentric perception for prototype wearable glasses!
Spatial computing empowered by augmented reality (AR) glasses emerges as a new computing paradigm. One of its most transformative capabilities is super memory – the superpower to accurately recall objects and people an individual has been paying attention to or interacting with in the physical world. In this work, we propose MemoLens that empowers AR glasses with super memory. Achieving such capability is not trivial: it requires storing vast amounts of visual data as compact digital memory and retrieving relevant information from the memory efficiently when being prompted. MemoLens achieves this via two key innovations. First, MemoLens leverages eye gaze information captured by the AR glasses to detect what the individual is paying attention to, and introduces a spatio-temporal token compression technique to generate highly compact visual memory. Second, MemoLens introduces a hierarchical retrieval architecture for fast and accurate multimodal memory recall.
@inproceedings{alam2026memolens,title={MemoLens: Empowering Augmented Reality Glasses with Super Memory},author={Alam, Samiul and Siam, Shakhrul Iman and Zhang, Mi},booktitle={Proceedings of the 24th Annual International Conference on Mobile Systems, Applications and Services (MobiSys)},pages={737--749},year={2026},doi={10.1145/3745756.3809235},}
AI glasses present a compelling platform for AI agents to serve as personalized memory assistants. To be genuinely useful, such systems must move beyond short-term video comprehension and address memory gaps that humans experience for practical, personal, or social purposes over longitudinal egocentric video streams. However, existing egocentric datasets predominantly focus on action recognition or generic QAs from short clips, measuring perceptual capabilities rather than realistic human memory needs. We introduce SuperMemory-VQA, an egocentric visual question answering (VQA) dataset for evaluating AI assistants on practical, long-horizon memory tasks. It contains 52.9 hours of everyday activities recorded with AI glasses, including synchronized RGB video, audio transcription, eye gaze, IMU, and SLAM trajectories. Through a human-verified annotation pipeline, we construct 4,853 grounded question-answer pairs that span object and location memory, intent recall, visual scene recall, timeline reconstruction, conversational memory, and in-context retrieval. Each question is posed as multiple-choice with an explicit unanswerable option to test hallucination robustness. Benchmarking leading agentic frameworks and LLM backbones reveals that existing systems remain far from reliable on real-world memory tasks, highlighting the need for new architectures for grounded AI memory that can answer only when evidence is sufficient.
@article{alam2026supermemory,title={SuperMemory-VQA: An Egocentric Visual Question-Answering Benchmark for Long-Horizon Memory},author={Alam, Samiul and Siam, Shakhrul Iman and Proulx, Michael J. and Fort, James and Newcombe, Richard and Kim, Hyo Jin and Zhang, Mi},journal={arXiv preprint arXiv:2606.00825},year={2026},}
To enable egocentric contextual AI in always-on smart glasses, it is crucial to be able to keep a record of the user’s interactions with the world, including during reading. In this paper, we introduce a new task of reading recognition to determine when the user is reading. We first introduce the first-of-its-kind large-scale multimodal Reading in the Wild dataset, containing 100 hours of reading and non-reading videos in diverse and realistic scenarios. We then identify three modalities (egocentric RGB, eye gaze, head pose) that can be used to solve the task, and present a flexible transformer model that performs the task using these modalities, either individually or combined. We show that these modalities are relevant and complementary to the task, and investigate how to efficiently and effectively encode each modality. Additionally, we show the usefulness of this dataset towards classifying types of reading, extending current reading understanding studies conducted in constrained settings to larger scale, diversity and realism.
@inproceedings{yang2025reading,title={Reading Recognition in the Wild},author={Yang, Charig and Alam, Samiul and Siam, Shakhrul Iman and Proulx, Michael J. and Mathias, Lambert and Somasundaram, Kiran and Pesqueira, Luis and Fort, James and Sheriffdeen, Sheroze and Parkhi, Omkar and Ren, Carl and Zhang, Mi and Chai, Yuning and Newcombe, Richard and Kim, Hyo Jin},booktitle={The Thirty-Ninth Annual Conference on Neural Information Processing Systems},year={2025},}
SVD-LLM V2: Optimizing Singular Value Truncation for Large Language Model Compression
Xin Wang, Samiul Alam, Zhongwei Wan, and 2 more authors
In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), 2025
Despite significant advancements, the practical deployment of Large Language Models (LLMs) is often hampered by their immense sizes, highlighting the need for effective compression techniques. Singular Value Decomposition (SVD) emerges as a promising method for compressing LLMs. However, existing SVD-based compression approaches suffer from substantial truncation losses, leading to severe performance degradation in compressed models. In this work, we introduce SVD-LLM V2, a novel SVD-based LLM compression method that optimizes singular value truncation in SVD compression with dynamic compression ratio allocation across layers and loss-optimized weight truncation.
@inproceedings{wang2025svdllmv2,title={SVD-LLM V2: Optimizing Singular Value Truncation for Large Language Model Compression},author={Wang, Xin and Alam, Samiul and Wan, Zhongwei and Shen, Hui and Zhang, Mi},booktitle={Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)},pages={4287--4296},year={2025},}
Awarded the 2024 IEEE Internet Computing Best Paper Award.
Equipped with sensing, networking, and computing capabilities, Internet of Things (IoT) such as smartphones, wearables, smart speakers, and household robots have been seamlessly weaved into our daily lives. Recent advancements in Generative AI exemplified by GPT, LLaMA, DALL-E, and Stable Diffusion hold immense promise to push IoT to the next level. In this article, we share our vision and views on the benefits that Generative AI brings to IoT, and discuss some of the most important applications of Generative AI in IoT-related domains. Fully harnessing Generative AI in IoT is a complex challenge. We identify some of the most critical challenges including high resource demands of the Generative AI models, prompt engineering, on-device inference, offloading, on-device fine-tuning, federated learning, security, as well as development tools and benchmarks, and discuss current gaps as well as promising opportunities on enabling Generative AI for IoT. We hope this article can inspire new research on IoT in the era of Generative AI.
@article{wang2024iot,title={The Internet of Things in the Era of Generative AI: Vision and Challenges},author={Wang, Xin and Wan, Zhongwei and Hekmati, Arvin and Zong, Mingyu and Alam, Samiul and Zhang, Mi and Krishnamachari, Bhaskar},journal={IEEE Internet Computing},volume={28},number={5},year={2024},}
Large Language Models (LLMs) have demonstrated remarkable capabilities in important tasks such as natural language understanding, language generation, and complex reasoning and have the potential to make a substantial impact on our society. Such capabilities, however, come with the considerable resources they demand, highlighting the strong need to develop effective techniques for addressing their efficiency challenges. In this survey, we provide a systematic and comprehensive review of efficient LLMs research. We organize the literature in a taxonomy consisting of three main categories, covering distinct yet interconnected efficient LLMs topics from model-centric, data-centric, and framework-centric perspective, respectively. We have also created a GitHub repository where we compile the papers featured in this survey at https://github.com/AIoT-MLSys-Lab/Efficient-LLMs-Survey, and will actively maintain this repository and incorporate new research as it emerges. We hope our survey can serve as a valuable resource to help researchers and practitioners gain a systematic understanding of the research developments in efficient LLMs and inspire them to contribute to this important and exciting field.
@article{wan2023efficient,title={Efficient Large Language Models: A Survey},author={Wan, Zhongwei and Wang, Xin and Liu, Che and Alam, Samiul and Zheng, Yu and Liu, Jiachen and Qu, Zhongnan and Yan, Shen and Zhu, Yi and Zhang, Quanlu and Chowdhury, Mosharaf and Zhang, Mi},journal={Transactions on Machine Learning Research},issn={2835-8856},year={2024},url={https://openreview.net/forum?id=bsCCJHbO8A},note={Survey Certification},doi={10.48550/arXiv.2312.03863},}
Most cross-device federated learning (FL) studies focus on the model-homogeneous setting where the global server model and local client models are identical. However, such constraint not only excludes low-end clients who would otherwise make unique contributions to model training but also restrains clients from training large models due to on-device resource bottlenecks. In this work, we propose FedRolex, a partial training (PT)-based approach that enables model-heterogeneous FL and can train a global server model larger than the largest client model. At its core, FedRolex employs a rolling sub-model extraction scheme that allows different parts of the global server model to be evenly trained, which mitigates the client drift induced by the inconsistency between individual client models and server model architectures. Empirically, we show that FedRolex outperforms state-of-the-art PT-based model-heterogeneous FL methods (e.g. Federated Dropout) and reduces the gap between model-heterogeneous and model-homogeneous FL, especially under the large-model large-dataset regime. In addition, we provide theoretical statistical analysis on its advantage over Federated Dropout. Lastly, we evaluate FedRolex on an emulated real-world device distribution to show that FedRolex can enhance the inclusiveness of FL and boost the performance of low-end devices that would otherwise not benefit from FL. Our code is available at: https://github.com/AIoT-MLSys-Lab/FedRolex.
@inproceedings{alam2022fedrolex,title={FedRolex: Model-Heterogeneous Federated Learning with Rolling Sub-Model Extraction},author={Alam, Samiul and Liu, Luyang and Yan, Ming and Zhang, Mi},booktitle={Advances in Neural Information Processing Systems},year={2022},}