Curriculum Vitae — Samiul Alam
Curriculum Vitae of Samiul Alam — Software Engineer & AI Researcher at Google, CSE Ph.D. from The Ohio State University, former Research Scientist Intern at Meta Reality Labs, and Co-Founder of Bengali.AI.
Summary
Samiul Alam is a Software Engineer and AI Researcher at Google (Mountain View, CA) with a Ph.D. in Computer Science and Engineering from The Ohio State University and prior Research Scientist experience at Meta Reality Labs. His research and engineering span spatial computing, supermemory for augmented reality smart glasses, egocentric computer vision, multimodal foundation models, agentic recommendation systems, machine learning systems, on-device AI, model compression, and federated learning.
Experience
-
Software Engineer, Shopping Personalization
GoogleWorking with the Shopping Personalization team on user activity modeling and development of agentic recommendation systems.
- User activity modeling for personalization
- Agentic recommendation systems
-
Research Scientist Intern, Sensors and Systems Research
Meta Reality LabsWorked with the Sensors and Systems Research team at Meta Reality Labs on energy-efficient egocentric perception for prototype wearable glasses.
- Developed and deployed a reinforcement learning-based dynamic camera capture model on prototype wearable glasses
- Optimized on-sensor capture to minimize energy consumption and maximize AI assistant performance
-
Software Engineering Intern, Shopping Personalization
GoogleWorked with the Shopping Personalization team on multimodal large language models (MLLMs) and synthetic-to-real user browsing data alignment for personalized recommendation systems.
- Multimodal LLMs for personalized recommendation systems
- Synthetic-to-real user data alignment
-
Graduate Research Associate & University Fellow
The Ohio State UniversityConducted doctoral research in the AIoT-MLSys Lab advised by Prof. Mi Zhang, in collaboration with the Surreal team at Meta Reality Labs, on personal supermemory for AR smart glasses, egocentric vision, and efficient LLMs.
- Collaborated with the Surreal team at Meta Reality Labs on selective memory capture, gaze-guided spatio-temporal compression and retrieval, and long-horizon egocentric visual reasoning for AR glasses
- Developed singular-value-truncation compression algorithms for efficient large language models (SVD-LLM V2, NAACL '25)
-
Graduate Research Assistant
Michigan State UniversityBuilt heterogeneity-aware federated learning systems; published across top venues.
- Geo-distributed and heterogeneous federated learning (FedRolex, FedAIoT, FedSEA, FedAudio)
- Publications at NeurIPS, SenSys, ICASSP, INTERSPEECH, and DMLR
-
Research Assistant
University of HoustonWorked on deconvolution, state-space modeling, and Bayesian filters.
- State-space models for physiological and cognitive signal analysis
- Bayesian filtering and sparse multichannel decomposition
-
Software Engineer (Engineer I, Mobile Applications)
Samsung R&D Institute BangladeshExtended BLE interfacing for wearable apps on iOS and Android; developed cloud service SDK; earned Samsung Software Certification (Top 10%).
- BLE interfacing for Samsung Galaxy Wearable apps
- Controller app for 5G smart router and cloud service SDK
- Samsung SWC Professional Certification (Top 10%)
-
Co-Founder
Bengali.AINon-profit for open-source Bengali datasets and benchmarks for speech, vision, and text.
- Community dataset curation (NumtaDB, Bengali Handwritten Graphemes, Common Voice Bengali, OOD-Speech)
- Organized international machine learning competitions on Kaggle
Education
-
Ph.D. in Computer Science and Engineering
The Ohio State UniversityGPA: 3.75
- Dissertation: “Towards Supermemory in Spatial Computing: A Multimodal Approach for Memory Creation, Retrieval, and Visual Reasoning”
- Advisor: Prof. Mi Zhang | Committee: Prof. Xueru Zhang, Prof. Zhihui Zhu
- Awarded the OSU Graduate School University Fellowship
-
M.S. in Computer Science
Michigan State UniversityGPA: 3.86
- Thesis: “Federated Learning Benchmarks and Frameworks for Artificial Intelligence of Things”
-
B.Sc. in Electrical and Electronic Engineering
Bangladesh University of Engineering and Technology (BUET)GPA: 3.67
Honors & Awards
-
Best Paper Award
IEEE Internet ComputingAwarded for “The Internet of Things in the Era of Generative AI: Vision and Challenges.”
-
Graduate School University Fellowship
The Ohio State UniversityAwarded the competitive OSU Graduate School University Fellowship.
-
Community Competition Creator Prize
KaggleFor organizing the DL Sprint speech transcription competition with BUET CSE.
-
SWC Professional Certification
Samsung ElectronicsEarned Samsung Software Certification (Top 10%) in data structures and algorithms.
-
Semifinalist, International Future Energy Challenge
IFEC 2017Designed a high-efficiency, high-density DC/DC converter.
-
Honorable Mention, IEEE Signal Processing Cup
IEEE SP Cup 2017Real-time musical beat tracking on Raspberry Pi.
-
Category Winner, IEEE Maker Project
IEEEBuilt an educational 3D scanner; winner in Education category.
Core Areas & Skills
Spatial Computing & Egocentric AI
Foundation & Large Language Models
Machine Learning Systems
Federated & Distributed Learning
Academic Service
-
Technical Program Committee (TPC) Member
IEEE International Conference on Mobility, Sensing and Networking (IEEE MSN 2026)Served on the Regular Tracks Program Committee reviewing mobile systems, sensing, and edge networking research.
-
Program Committee Member & Conference Reviewer
Top-Tier AI, Vision & Systems VenuesProgram Committee / Reviewer for ICLR 2027, AAAI 2027, CVPR 2027, ICASSP 2027, CIKM 2026 (LASS Workshop), NeurIPS, TMLR, and ACM TOSN.
Selected Systems & Benchmarks
-
MemoLens
Empowering augmented reality glasses with super memory via gaze-guided spatio-temporal token compression and hierarchical multimodal retrieval (ACM MobiSys '26).
- Gaze-aware token compression
- Hierarchical multimodal retrieval
-
SuperMemory-VQA
Long-horizon egocentric visual question-answering benchmark built from 52.9 hours of multi-sensor AI glasses recordings with 4,853 human-verified QA pairs.
- Longitudinal egocentric memory
- Hallucination-robust VQA evaluation
-
Reading Recognition in the Wild
First large-scale multimodal dataset (100 hours) and transformer model using egocentric RGB, eye gaze, and head pose for selective memory creation on smart glasses (NeurIPS '25).
- 100-hour multimodal dataset
- RGB + gaze + head pose fusion
-
SVD-LLM V2 Model Compression
Optimizing singular value truncation via dynamic compression ratio allocation and loss-optimized truncation for efficient LLM deployment (NAACL '25).
- SVD compression
- Dynamic rank allocation
-
FedRolex & FedAIoT
Frameworks and benchmarks for model-heterogeneous and authentic AIoT federated learning under varied bandwidth and compute constraints (NeurIPS '22, DMLR '24).
- Rolling sub-model extraction
- 8 authentic IoT device datasets
-
5G Smart Router Interfacing
Interface and AR visualization for mesh network strength, channel estimation, and optimal router placement.
- 5G router
- AR visualization