Daniel Lobaton
Research Engineer at Mechanize — alignment & RL environments for frontier models
I build RL environments and alignment evals for frontier models, and use them to study how models learn — and whether they internalize what we intended rather than what merely scores well under evaluation. My focus is measurement I can trust.
Experience
-
Mechanize Now Research Engineer
High fidelity RL environments and alignment evals for frontier labs — building the tasks, some reward functions, verified, and sorrouninding infra.
-
G2X Founding Engineer
Founding engineer on a federal procurement platform. Across separate engagements: led fraud investigations with the FBI, redesigned legacy data systems at CMS, and briefed the Coast Guard. Won an AFFIRM award for my work.
-
Microsoft Software Engineer
Profiled and optimized power consumption for Copilot across a billion-device fleet — instrumenting real-world power draw and validating the gains before launch.
-
Tesla Research Engineer, Intern
Research engineer; Manufactoring OS team, drift & anomaly detection in manufacturing line.
-
University of Florida B.S. Computer Science
Undergraduate research at the SimLab, building medical simulations to train battlefield doctors.
Writing
- The second wave of PPO Draft
Essays and research notes — in progress.
Elsewhere
- Email dlobaton2@gmail.com
- GitHub github.com/danlobaton
- LinkedIn linkedin.com/in/danlobaton