Stellamaris (W) Nakacwa
My Profile
Name: Stellamaris Nakacwa
Location: USA
Capacity: Research | Machine Learning Systems | Earth Systems | Technical Leadership
LinkedIn: stellawava
Contact: email
Google Scholar: Profile
About Me
I am a research scientist and engineer [ML/AI] focused on shaping intelligent systems and approaches with important data and algorithmic solutions. I earned this with over a decade of practice, learning, and applying myself to various knowledge environments, and creating strategies and solutions to some of the world's prevailing challenges.
For nearly a decade, I have moved from mapping places applying Geomatics, GIS & Earth Observations(EO) to leading such programs and then developing Artificial Intelligence architecture, applying custom techniques to inference and optimize various models for the US National Energy Research, contributing to various research goals in the Genesis Mission.
I am very curious about systems, and when I'm not doing research, I'm possibly working on some interesting ideas in my writing lab - Stella_Space, or reading/watching screen writes about knowledge production & governance systems of a given century.
I am building a simple book and screen write recommendation system here: [OOO]
My Training
I obtained my MSc. Scientific Computing at Harrisburg University with Prof. Majid's guidance, a Master's in Geography at West Virginia University and direct supervision from Prof. McCusker, and a BSc. degree from Makerere University under mentorship of Prof. Gidudu.
I am also a Leadership Fellow for YouthMappers cohort 2017, and won numerous awards for my community and YouthMappers at large taking on various geo-programs.
Surveying/Geomatics grounded me with engineering concepts and mathematical pieces that continue to drive my interests across EO and High Performance computing systems today.
Selected Projects
Data Movement - Explaining & implement various data movement algorithms and GPU profiling and benchmarking [code][blog]
LLMs - Transformer architecture to train AI models exploring FLashAttention & Triton Optimization [code] [code]
Document Intelligence - Both transformer and non-transformer approach for pdf data processing. [code][Product]
CompSys - A matrix cost visualizer for gpu performance mapping for different matrix layouts. [code]
CMM - A large cross teams project based out of the Critical Minerals Center of Excellence focused on securing critical minerals supply chains through research and technology development. Data team and lead for advanced EO & computing methods for unconventional discovery. Read: Phase Presentation
Rokbase Core Facility - The largest environment for geological core images across all 50 U.S. states. [Insights]
WELLBASE - Large Model Finetuned Wellbore Material Assessment Application. [Presentation]
NGI: A large cross lab pipeline Integration with Optical Sensors: project at NETL to improve natural gas and hydrogen pipelines safety. https://netl.doe.gov/node/15505
Spotlight Kampala - An inter-university collaboration doing energy research for access, affordability, reliability, and clean cooking solutions. [Website]
Crop Mapping - A project applying computer visions to identify farmlands and quantify crop types [Blog] [Blog]
Experience
Research Scientist-AI/ML - National Energy Technology Laboratory(NETL) | Leidos
Program Director - YouthMappers
SME Synthetic Data - Handshake AI
Graduate Research Assistant - West Virginia University
Geospatial Scientist - Landnet Uganda
Writing - Featured Publications | Conferences | Stories
Publications
- Lite Learning: Efficient Crop Classification in Tanzania Using Feature Extraction with Machine Learning & Crowd Sourcing 2026
- Dynamic-Query Robustness of ANN Indexes under Time-Indexed Drift 2026
- A National Critical Minerals and Materials Prospectus and Resource Assessment Methodology 2025
- RokBase Core Facility Data Collection 2024
- WELLBASE-An Interactive Platform for Wellbore Material Assessment 2024
- Open mapping towards sustainable development goals: voices of youthmappers on community engaged scholarship 2023
- Carbon Storage Wellbore Materials Data Tool 2023
- Rokbase: Digital Rock Visualization and Exploration Web Application 2023
- Cities of the Future Need to Be Both Smart and Just: How We Think Open Mapping Can Help 2022
- Combining Volunteered Geographic Information and WPdx standards to Improve Mapping of Rural Water Infrastructure in Uganda 2022
- Digital humanitarians for the sustainable development goals: YouthMappers as a hybrid movement 2022
Stories
Changes in scientific needs, and the 3Vs of data, have produced an incredible shift in the architecture of computing we have today. Parallelism remains a fundamental computing foundation influencing our decisions on algorithms and scaling decisions around our architecture. My goal is to bring you up to speed by exploring many algorithms on today's most powerful scientific processors: GPUs, through Quinn Michael's PCAM model. [Visit Site]
How else would you learn about advanced computing and its value to the scientific domain? This is a storytelling approach to scientific computing, starting with the term "Super Computer" in the 1940s. My goal is for you to learn about the energy challenges by understanding the persistent challenges facing AI computing today. [Substack] Youtube: TBD