Research engineer at Anduril, autonomy and wargaming. Occasional token arsonist. Seattle.
I build the tooling that lets language models author simulated worlds, and I get agents to make real decisions inside them. The thread running through most of my work is agent orchestration: many agents doing real work together, reliably, in environments that push back.
I came to this through math. I studied applied math at CU Boulder, spent a few years modeling missions and autonomy in defense simulation, and kept circling the same two questions. How does behavior emerge from simple rules, and how do you build a model you can actually trust when the agents start changing themselves.
Outside work I build small things in the open, mostly around agents, dynamical systems, and markets. Some of them are below. Friends call me Dani.
A short introduction. Opens on YouTube.
Now
At Anduril: tooling for language-model authoring of scenarios and combined physics and autonomy models, natural language tasking for robots in simulation, and a library of behavior models I can point at specific problems.
On my own time: spawning agents into generated game worlds, then building up a library of scenarios and behaviors for them. It is the most fun I have with a computer.
Reading toward modern machine learning systems, and the economics underneath all of it.
Updated June 2026.
Work
Anduril Industries
Research Engineer, Autonomy and Wargaming. 2026 to present, Seattle.
Built and shipped tooling for language-model authoring of simulated scenarios and combined multiphysics and autonomy models, used across the organization.
Led a zero-to-one natural language tasking feature for robots in a simulated environment.
Built and maintained a library of high-fidelity robot behavior models, and used them to develop concepts of operations for specific customers.
LinQuest
Modeling and simulation engineering. 2024 to 2025, Washington DC.
Wrote a C++ plugin for the AFSIM engine that extended its finite-state-machine features and modularized complex behavior modeling.
Modeled missions and kill chains, including ISR, SEAD, and electronic attack, working with subject matter experts to capture concepts of operations.
Built Python tooling to batch-run large wargaming experiments and turn mission output into dashboards and visualizations.
Built a RAG-based, quantized LLM command-line tool for querying AFSIM documentation, and wrote a whitepaper on using language models for modeling-script generation.
Earlier
Research at the National Science Foundation (5G traffic anonymization, nonnegative matrix factorization for secure communications) and the Air Force Research Laboratory (reinforcement-learning agents in AFSIM, multi-asset control policies for multi-ship and multi-UAV scenarios).
BS Applied Math, University of Colorado Boulder, 2024. Cleared US SECRET.
