Grace Fokam

Grace Fokam

AI Researcher & Climate Technologist

const researcher = {
focus: "Generalizable AI for the physical world",
building: "Fovra"
};

>About Me

Lately, I've been building Fovra, an AI research company exploring how machines can learn to understand, predict, and optimize the physical world. I'm Cameroonian-American, grew up in Maryland, and am currently based in the Washington, DC area.

>Research Interests

  • Physical AI
  • World Models
  • Representation Learning
  • Computer Vision
  • Geospatial AI

>Ventures

>Research

Google Research & Brown CS

Built a modular Python ML pipeline for surrogate modeling, evaluation, and search optimization, achieving 76.8% explained variance and 68% surrogate-guided search success.

NASA Goddard

Build CNN-based crop-yield models and geospatial ML pipelines using Sentinel-2, Landsat-8, and 50K+ field observations, reducing preprocessing time by 30%.

Massive Data Institute at Georgetown

Fine-tuned a BERT-based model on 33K+ web data points to predict forced displacement during the Sudan Civil War, achieving 81% accuracy.

Extreme Materials Institute at Johns Hopkins

Developed a PyTorch and OpenCV computer-vision model to reconstruct steel microstructures with 91% accuracy.

>Selected Recognitions

>Speaking

>Connect