Clay Foye

clay.foye@gmail.com | Github

Research Interests

Mechanistic interpretability · Digital Humanities · Visual Studies

Currently investigating how multimodal transformers encode and manipulate cross-modal concepts, and how those concepts fit in larger visual traditions. My goal is to connect internal model mechanisms with researchers in the humanities and develop interpretable model architectures and tools.

Education

École Polytechnique Fédérale de Lausanne (EPFL) Lausanne, Switzerland

MSc in Digital Humanities (GPA 5.45/6) Present, expected end August 2026

  • Master's Thesis: Natural Language Reasoning on Geospatial Knowledge Graphs (in progress)
  • Semester Project: Biology of a VLM: Transcoder Training for Multimodal Interpretability
  • Relevant Courses: Machine Learning for DH · Computational Social Media · Applied Data Analytics

Dartmouth College Hanover, NH

B.A. in Computer Science and Middle Eastern Studies (GPA 3.64/4.0) 2022

  • CS Thesis: "A Logical Re-conception of Neural Networks: Hamiltonian Bitwise Part-Whole Architecture"
  • MES Honors Thesis: Hacking Orientalism: Theorizing Arabic Digital Writing
  • Honors: Third Honor Group (top 15%), 2021–2022

Work Experience

Odoma Ltd - Thesis Project Intern Mar 2026 — Aug 2026

  • Researching agentic and skill-based architectures for interacting with knowledge graphs
  • Developing benchmarks and evals for OpenClaw-style system deployments

University of Zurich: Research Assistant Sep 2025 — Mar 2026

Digital Visual Studies Lab Zurich, Switzerland

  • Investigated circuit-level interpretability in transformer architectures
  • Analyzed city-scale point-cloud data for spatial and sightline modeling
  • Contributed to DVS conference submissions and projects

Creare LLC - Engineer Jan 2023 — Jan 2024

  • Developed ML pipelines for point-cloud data, robotics control systems, and geospatial APIs
  • Co-authored the company's first AI usage guidelines linking technical development to ethical and legal frameworks
  • Prototyped LLM-based agent integration for enterprise data retrieval (LangChain, OpenAI API)

Machine Learning Technician Jan 2022 — Jan 2023

Citrin Family GIS / Spatial Analysis Lab, Dartmouth College Hanover, NH

  • Applied ML methods in Julia and ArcGIS to large geospatial datasets for spatial analytics

Teaching Assistant 2021

Dartmouth College – Computer Science Department Hanover, NH

  • Led office hours and group debugging sessions for introductory programming courses

Conference Presentations

  • "The Finite Potentials of Latent Spaces." With Paul Guhennec. Presentation at From Hype to Reality Symposium, University of Zurich, 2025.
  • "Navigating Institutional Archives of Suisse Romande Architecture Competitions." With Paul Guhennec. Long paper at Historical Network Research Conference, Turin, July 2026.
  • "Architectural Networks and Institutional Critique: Suisse Romande Competition through the Lens of Mathematical Graphs." With Paul Guhennec. Long paper at Networks in Architecture and Architectural History, Braunschweig, September 2026.
  • "The Latent Space Trial." With Dario Neguereula del Castillo, Ludovica Schaerf, Paul Guhennec, Iacopo Neri. Performance at xCoAx, Turin, 2026.
  • "Digital Throws of the Dice: A Genealogy of Rhetoric in Fallout." Paper at Transmedia History: Circulations, Reconfigurations and New Methodologies, University of Lausanne, January 2025.

Projects

EPFL, Mechanistic Interpretability Tools

  • Built cross-layer transcoders from scratch for multimodal transformers using PyTorch and HuggingFace to identify and steer concept circuits

Creare, Point-Cloud Reconstruction

  • Converted raw 3D scans into CAD models using OpenVDB and C++

Creare, Open-source development of PODPAC

  • Developed Python pipelines for large-scale meteorological data processing

Skills

  • Languages & Frameworks: Python, C, Rust, Julia, PyTorch, TensorFlow, HuggingFace, NumPy
  • Tools & Platforms: Git, Docker, OpenClaw, ROS, OpenVDB, ArcGIS