Experience

Applied Scientist Intern

Amazon, GRAISE teamSeattle, WA

Multimodal AI for retail smart glasses: real-time on-device vision, cloud foundation models (VLMs, LLMs, SAM), and hands-free AR and voice workflows.

Graduate Research Intern (GRO program), GeoAI Group

Oak Ridge National LaboratoryOak Ridge, TN

AI-assisted characterization of post-Hurricane Helene forest damage from UAS imagery.

• Adapted foundation models (SAM, Mask2Former) and fused LiDAR- and photogrammetry-derived height products for instance-level detection of fallen logs.
• Built the CWDBench benchmark (WACV 2027 Datasets Track): 10k+ verified instances with oriented boxes and masks, cross-site and cross-sensor evaluation, and a centerline metric for elongated objects.
• Co-developed ARFA, an agentic riverine-flood-response assistant selected as a finalist in the ACM SIGSPATIAL OASIS 2026 challenge.

Graduate Research Assistant

Biocomplexity Institute, University of VirginiaCharlottesville, VA

Knowledge-guided computer vision and large-scale datasets for agricultural and environmental mapping, advised by Madhav Marathe.

• Released IrrMap (KDD 2025) and IRRISIGHT (NeurIPS 2025), two of the largest public multimodal datasets for irrigation and agricultural water management, with end-to-end ML-ready pipelines.
• Designed KIIM (IJCAI 2025), a knowledge-guided irrigation-mapping model that injects crop-type and land-use priors through cross-attention, and IGraSS (IJCAI 2025), which couples segmentation with graph constraints for canal-network extraction.
• Built the PRISM-CAFO (WACV 2026) and CAFOSat (EarthVision @ CVPR 2026) pipelines for prior-conditioned livestock-facility detection and strongly annotated dataset creation from NAIP imagery.

Python, PyTorch, Hugging Face, OpenCV, rasterio/GDAL, QGIS, scikit-learn, NumPy, pandas, Matplotlib

Research Intern

USDA-ARS (SCINet / AI-COE Graduate Student Internship Program)Remote

Irrigation mapping from multispectral Landsat imagery.

• Built geospatial training sets and evaluated spectral-index and band combinations for irrigation classification.
• Developed IrrNet (V4A @ CVPR 2024), an incremental patch-size training scheme that improved accuracy by roughly 10% over prior baselines.

Software Engineer

Enosis SolutionsDhaka, Bangladesh

Full-stack development on production systems in a 6–8 person team.

• Built an incident-management tool in C#/MVC and a web platform for training and testing machine-learning models (React, Node.js, SQL/MongoDB).

C#, JavaScript, React, Node.js, Express, MySQL, Oracle, MongoDB, Git, Jira

Undergraduate Researcher

Ahsanullah University of Science & TechnologyDhaka, Bangladesh

Real-time sign-language recognition and document understanding with deep learning.

• Created the BdSL36 dataset and real-time detectors for Bangladeshi Sign Language (ICIET 2018, MLCSA @ ACCV 2020).

Undergraduate Research Mentor

Biocomplexity Institute, University of Virginia

Mentoring undergraduate researchers in computer vision and remote sensing: dataset preparation, model evaluation, reproducibility, and paper writing.

Graduate Teaching Assistant

Department of Computer Science, University of Virginia

Neural Networks (Spring 2025), Digital Signal Processing (Spring 2024), Machine Learning (Spring 2023), Mobile Application Development (Fall 2022).

Ph.D. in Computer Science

University of Virginia

• Advisor: Madhav Marathe, Biocomplexity Institute
• UVA SEAS Endowed Fellowship (2025); UVA CS PhD Fellowship (2021)
• Focus: knowledge-guided machine learning, multimodal understanding, vision–language models, remote sensing

M.S. in Computer Science

University of Virginia

• Coursework: machine learning, ML for image processing, interpretable ML, NLP, 3D computer vision, algorithms, software security

B.Sc. in Computer Science and Engineering

Ahsanullah University of Science & Technology

• Dean’s List of Honor; CGPA 3.79 / 4.00
• Thesis: Detecting Letters and Words from Bangladeshi Sign Language in Real-Time (thesis, slides); supervisor Mohammad Imrul Jubair