MATH
Model behavior
Mechanistic interpretability, representation engineering, experimental design, and careful language around what an observed result does—and does not—show.
About
Research, systems, and strange interfaces
I’m Corina Kaiser, an NYC-based research engineer with an MS in Mathematical Sciences from NYU Tandon. I build AI systems and research tools that make model behavior inspectable, usable, and operational—from multi-agent memory and real-time applications to representation-engineering experiments and physical interfaces.
My practice sits between mathematical investigation and product engineering. I care about what a system does, how we can measure it, and what it takes to put it in someone’s hands responsibly. That has led me to work across retrieval and agent memory, full-stack collaboration tools, interactive data visualization, model-routing research, and embedded interfaces.
I began this path through experimental dialogue systems and still value the unusual, generative questions that brought me there. brain.vat ↗ remains an early example: a multi-agent dialogue experiment built around autonomous GPT-2 interaction.
Practice
SYS
Systems in use
Real-time applications, semantic retrieval, services, interfaces, and operational tools designed around actual people and workflows.
Outside the terminal
I have played rugby for more than a decade, including the 2025 season with New York Exiles, and make work across painting, collage, film photography, and clothing design. Those practices inform how I build: materially, experimentally, and with attention to the person encountering the system.