Machine Learning Research Engineer (Hybrid Eligible)

Requisition Id 14225 

 

Overview: 

As a U.S. Department of Energy (DOE) Office of Science national laboratory, ORNL has an extraordinary 80-year history of solving the nation’s biggest problems. We have a dedicated and creative staff of over 6,000 people! Our vision for diversity, equity, inclusion, and accessibility (DEIA) is to cultivate an environment and practices that foster diversity in ideas and in the people across the organization, as well as to ensure ORNL is recognized as a workplace of choice. These elements are critical for enabling the execution of ORNL’s broader mission to accelerate scientific discoveries and their translation into energy, environment, and security solutions for the nation.

 

We are seeking a Machine Learning (ML) Research Engineer who will support the development of self-supervised learning methods for large vision-language models to benefit downstream tasks including object detection and counting, visual question answering, semantic segmentation, change detection and polygonization of geospatial vector geometries across research projects at ORNL.  This position resides in the GeoAI Research Group in the Geographic Data Science Section, Geospatial Science and Human Security (GSHS) Division, National Security Sciences Directorate, at ORNL.

 

As part of our team, you will support research tasks related to optimizing codes for scaling foundation models training, fine-tuning to several downstream tasks and lead polygonization of building footprint vector geometries. The group conducts cutting edge research and publishes on novel ML-based solutions to large scale geospatial application challenges. Research activities include the design of efficient ML workflows using high performance computing (HPC) environments, preprocessing and transforming large volumes of satellite imagery, and conducting post-processing and validations of model of outcomes. Under the guidance of senior research scientists, the selected applicant will take roles on multidisciplinary teams supporting ground breaking research and engineering with large-scale distributed ML workflows, using ORNL’s Frontier exascale supercomputer for its dense GPU-based HPC resources to train large GeoAI models.

 

Major Duties/Responsibilities: 

• Develop and implement new methods for polygonization of building footprint vector geometries.

• Develop and implement workflows to support large scale self-supervising learning for large vision-language models.

• Collect, process, and analyze large volumes of satellite imagery

• Support the design and implementation of efficient foundation models finetuning methods.

• Visualize and communicate analysis results via technical reports, and peer-reviewed publications.

• Collaborate with other research and technical professionals on new methods to advance GeoAI for end-to-end multi-modality geospatial data analytics.

• Deliver strong science and engineering artifacts demonstrating research innovation for our sponsors.

• All team members deliver ORNL’s mission by aligning behaviors, priorities, and interactions with our core values of Impact, Integrity, Teamwork, Safety, and Service. Promote diversity, equity, inclusion, and accessibility by fostering a respectful workplace – in how we treat one another, work together, and measure success.

 

Basic Qualifications:

• MS in electrical engineering, civil engineering, geoinformation science, or a related field and two (2) years of applied experience (professional or academic lab setting). An equivalent combination of education and experience may be considered.

• Hands-on experience training machine learning models on HPC infrastructures using GPU accelerators.

• Experience building data-fusion workflows to ingest multi-modality geospatial data.

• Experience using Python or other programming languages to develop AI algorithms in PyTorch computing framework.