Lead Machine Learning Engineer & Head of Insights at GatherGov • MSCS at NYU Courant • BSAM at Northwestern
Hello! I'm Alexandre Kaiser, Lead Machine Learning Engineer & Head of Insights at GatherGov. I design and build AI systems that make local government data across the United States more transparent and actionable.
In 2024, I earned my Master's in Computer Science at NYU Courant, where I conducted research on deep learning theory under Arthur Jacot. Prior to that, I graudated with a Bachelor's in Applied Mathematics from Northwestern University—driven by a curiosity in how mathematical principles impact each domain.
Raised in a business-minded family, but drawn to philosophical questions, I thrive at the intersection of theory and real-world impact.
My work focuses on building scalable ML systems that bridge theoretical understanding with real-world impact. At GatherGov, I lead the full lifecycle of ML-powered insights, from strategic planning to hands-on development, delivering actionable intelligence through state-of-the-art models and agentic workflows.
My outward goal is to forge a dynamic balance between innovation and wastefulness: to relentlessly generate new solutions for problems others find worth solving, marshaling every resource to its fullest—wringing water from stones in the pursuit of value creation.
My inward goal is to cultivate the capacity to find happiness in everything; believing that beauty lies in the eye of the beholder and that any taste can be acquired, I see every moment of aversion as a missed opportunity to discover something truly beautiful.
GatherGov
NYU Courant - Arthur Jacot Lab
Neuron7
Prophesee
My research focuses on the theoretical foundations of deep learning, particularly in understanding how optimization affects feature learning and neural network dynamics. I investigate the implicit biases that emerge during training and their implications for model behavior and generalization.
CPAL 2025 (Oral Presentation) • Second Author
Proved the low dimensional bias of regularized neural networks through Hamiltonian mechanics framework.
View on arXivMaster's Thesis • NYU Courant
Investigated continual learning dynamics through theoretical analysis of deep linear networks.
View PDFResearch Survey
Surveyed proposal to consolidate optimal OCO algorithms across all convex geometries.
View PDFResearch Survey
Explored how Blackwell's Approachability Theory provides foundations for machine learning.
View PDFResearch Study
Explored reasons for Adam's empirical dominance by proving dynamical properties across 4 key regimes.
View PDFOpen Source Project
Created app to convert long-form texts into hierarchical relational graphs for summarization and search.
View on GitHub