cv

Software Engineering Intern

Google

Aligned synthetic data from LLMs to real user data to improve personalized recommender systems.

• Synthetic-to-real data alignment
• Personalization for recommenders

Graduate Research Associate

Ohio State University

Research on on-device AI, model compression and quantization, and efficient retrieval with deep hashing for multimodal LLMs; collaboration with Meta Surreal.

• Model compression and quantization for edge
• Deep hashing retrieval for efficient similarity search
• Multimodal LLM efficiency

Research Assistant

Michigan State University

Built heterogeneity-aware federated learning systems; published across top venues.

• Geo-distributed and heterogeneous FL
• Publications at NeurIPS, SenSys, ICASSP, InterSpeech, DMLR

Research Assistant

University of Houston

Worked on deconvolution, state-space modeling, and Bayesian filters.

• State-space models
• Bayesian filtering

Engineer I, Mobile Applications

Samsung R&D Institute Bangladesh

Extended BLE interfacing for wearable apps on iOS and Android; developed cloud service SDK; earned Samsung Software Certification (Top 10%).

• BLE interfacing for wearables
• Cloud service SDK
• Top 10% developer certification

Co-Founder

Bengali.AI

Non-profit for open-source Bengali datasets and benchmarks for speech and text.

• Community dataset curation
• Benchmarks for Bengali speech and text

PhD

The Ohio State University

Computer Science and Engineering

MS

Michigan State University

Computer Science

B.Sc.

Bangladesh University of Engineering and Technology (BUET)

Electrical and Electronic Engineering

Best Paper Award

IEEE Internet Computing

Awarded for “The Internet of Things in the Era of Generative AI: Vision and Challenges.”

Community Competition Creator Prize

Kaggle

For organizing the DL Sprint speech transcription competition with BUET CSE.

Semifinalist, International Future Energy Challenge

IFEC 2017

Designed a high-efficiency, high-density DC/DC converter.

Honorable Mention, IEEE Signal Processing Cup

IEEE SP Cup 2017

Real-time musical beat tracking on Raspberry Pi.

Category Winner, IEEE Maker Project

IEEE

Built an educational 3D scanner; winner in Education category.

LLM-based AR Assistant

On-device retriever indexes AR device data and retrieves segments for a cloud multimodal LLM.

• Edge retrieval
• Multimodal context

Aria Eye Tracking Dataset

Multimodal benchmark and dataset for eye-tracking-enhanced video understanding.

• Eye gaze + video
• Benchmarking

SVD-LLM Model Compression

SVD-based compression for Jetson and mobile using ONNX Runtime, MLC LLM, and TVM.

• SVD compression
• Edge deployment

Federated Learning for Heterogeneous Edge Devices

Frameworks for model-heterogeneous FL under varied bandwidth and compute constraints.

• Model heterogeneity
• Edge constraints

5G Smart Router Interfacing

Interface and AR visualization for mesh network strength and optimal placement.

• 5G router
• AR visualization