Geospatial Computer Vision and GenAI AI/ML at AWS

From 2020-2024, I worked as a Data Scientist and Machine Learning Engineer at AWS Worldwide Public Sector, assisting government and nonprofits with AI/ML solutions.

I collaborated with talented colleagues and customers on diverse use cases spanning various industries, technologies, and disciplines to train and deploy AI/ML models and pipelines.

Beyond leading and supporting delivery of successful award-winning project deliveries, I created reusable artifacts, prospected new POCs, and led initiatives that brought new work into our group.

My main AI/ML focus areas included:

  • Object classification, detection, and tracking from video, sound, and satellite imagery (RGB, MSI, SAR)
  • Generative AI models (LLM, RAG, Bedrock)
  • Self supervised and semi-supervised training (to provide classification lift)

Applications included:

My public contributions include model and processing support for an open-source solution for AI/ML processing of satellite imagery (overview  src: osml and osml-models).

Similar customer workflows with public documentation include damage assessment, crop segmentation, and video summarization.

I received an AWS Innovator of the Quarter award 2021, Builder Award 2022 for my contributions.

Tools: PyTorch, MXNet, SOTA model selection, training, data pre-processing, parallelization, model training and fine tuning, CUDA, runtime optimization, Arduino, Python, JupyterLab, SAMGeo

AWS Stack: CloudFormation/CDK, Lambda, SageMaker BYOC, S3, Redshift, Snowball edge, Bedrock, GovCloud


a subset of what we explored and used :


these public datasets were useful for prototyping, pre-training, and evaluating computer vision models.

MOT Challenge PET and MOT17