Aggrey Muhebwa

Seattle, WA

muhebwa.grey [at] gmail.com

I am an Applied Scientist at Amazon, where I build production NLP systems that transform open-source signals into decision-ready intelligence at enterprise scale. My work spans the full pipeline — from large-scale, multilingual ingestion across heterogeneous sources, through LLM-based relevance screening, fine-tuned domain-adapted classifiers, and RAG-based entity resolution that map unstructured events onto enterprise knowledge bases. I pair this with cross-model agreement, feature-attribution analysis, and systematic failure-mode documentation to keep signal quality high enough for downstream stakeholders to act on. The throughline is improving the timeliness, precision, and cost efficiency of monitoring at scale.

Before Amazon, I was a Postdoctoral Researcher at Stanford University with Prof. Khalid Osman, where I was awarded a $100K Precourt Institute grant to develop frameworks for principled accuracy-vs-emissions tradeoffs in AI deployment. I completed my PhD in Electrical and Computer Engineering at the University of Massachusetts Amherst under Prof. Jay Taneja, as a UN-IPCC Scholar (2021–2023) and e-GUIDE fellow. I hold an MS in ECE from Carnegie Mellon University and a BS in Computer Engineering from Makerere University.

My research sits at the intersection of applied ML, responsible AI, and large-scale infrastructure. I work on production LLM pipelines, model interpretability, energy-aware AI, and federated learning optimization — with a recurring throughline: making ML systems trustworthy enough to deploy in high-stakes domains where the cost of a wrong answer is real. For a full list of publications, see my Google Scholar profile (h-index 7, 260+ citations).

news

Aug 01, 2025	Started as Applied Scientist at Amazon, building production LLM systems for sustainability and responsible-AI deployment.
Apr 01, 2025	Awarded a $100K Bits & Watts Initiative grant (co-PI) for research on socio-psychological perspectives in AI model selection for performance and sustainability.
Feb 01, 2025	Paper accepted: A Behavioral Finance Framework for Balancing AI Accuracy and Operational Carbon Emissions — ACM Journal on Computing and Sustainable Societies.

selected publications

Preprint

Kuramoto-FedAvg: Using Synchronization Dynamics to Improve Federated Learning Optimization under Statistical Heterogeneity

Aggrey Muhebwa, Khotso Selialia, Fatima Anwar, and 1 more author

2025

Bib

@misc{muhebwa2025kuramoto,
  title = {Kuramoto-FedAvg: Using Synchronization Dynamics to Improve Federated Learning Optimization under Statistical Heterogeneity},
  author = {Muhebwa, Aggrey and Selialia, Khotso and Anwar, Fatima and Osman, Khalid K.},
  year = {2025},
  howpublished = {arXiv preprint},
}

Preprint

Causal Distillation: Transferring Structured Explanations from Large to Compact Language Models

Aggrey Muhebwa and Khalid K. Osman

2025

Bib

@misc{muhebwa2025causal,
  title = {Causal Distillation: Transferring Structured Explanations from Large to Compact Language Models},
  author = {Muhebwa, Aggrey and Osman, Khalid K.},
  year = {2025},
  howpublished = {arXiv preprint},
}

ACM JCSS

A Behavioral Finance Framework for Balancing AI Accuracy and Operational Carbon Emissions

Aggrey Muhebwa and Khalid K. Osman

ACM Journal on Computing and Sustainable Societies, 2025

Bib

@article{muhebwa2025behavioral,
  title = {A Behavioral Finance Framework for Balancing AI Accuracy and Operational Carbon Emissions},
  author = {Muhebwa, Aggrey and Osman, Khalid K.},
  journal = {ACM Journal on Computing and Sustainable Societies},
  year = {2025},
}

J. Comput. Civ. Eng.

Using Large Language Models for Systematic Failure Model Identification in Levee Infrastructure

Aggrey Muhebwa and Khalid K. Osman

Journal of Computing in Civil Engineering, 2025

Bib

@article{muhebwa2025levee,
  title = {Using Large Language Models for Systematic Failure Model Identification in Levee Infrastructure},
  author = {Muhebwa, Aggrey and Osman, Khalid K.},
  journal = {Journal of Computing in Civil Engineering},
  year = {2025},
}

WRR

Improving Discharge Predictions in Ungauged Basins: Harnessing the Power of Disaggregated Data Modeling and Machine Learning

Aggrey Muhebwa, C. J. Gleason, Dongmei Feng, and 1 more author

Water Resources Research, 2024

Bib

@article{muhebwa2024discharge,
  title = {Improving Discharge Predictions in Ungauged Basins: Harnessing the Power of Disaggregated Data Modeling and Machine Learning},
  author = {Muhebwa, Aggrey and Gleason, C. J. and Feng, Dongmei and Taneja, Jay},
  journal = {Water Resources Research},
  year = {2024},
}

J. Comput. Civ. Eng.

Promoting Equitable Access to Transboundary Hydrology Data: A Framework for Secure Sharing and Collaborative Machine Learning to Enhance River Discharge Prediction

Aggrey Muhebwa and Khalid K. Osman

Journal of Computing in Civil Engineering, 2024

Bib

@article{muhebwa2024transboundary,
  title = {Promoting Equitable Access to Transboundary Hydrology Data: A Framework for Secure Sharing and Collaborative Machine Learning to Enhance River Discharge Prediction},
  author = {Muhebwa, Aggrey and Osman, Khalid K.},
  journal = {Journal of Computing in Civil Engineering},
  year = {2024},
}

ACM JCSS

Pixel Perfect: Using Vision Transformers to Improve Road Quality Predictions from Medium Resolution and Heterogeneous Satellite Imagery

Aggrey Muhebwa, Gabriel Cadamuro, and Jay Taneja

ACM Journal on Computing and Sustainable Societies, 2023

Bib

@article{muhebwa2023pixelperfect,
  title = {Pixel Perfect: Using Vision Transformers to Improve Road Quality Predictions from Medium Resolution and Heterogeneous Satellite Imagery},
  author = {Muhebwa, Aggrey and Cadamuro, Gabriel and Taneja, Jay},
  journal = {ACM Journal on Computing and Sustainable Societies},
  year = {2023},
}