EB Eric Brahms
Résumé Get in touch
Principal Software Engineer · Technical Leader · XR, Robotics & Aerospace

Complex systems don't get to say they work,
they have to prove it.

I lead the teams that build that proof: robotics-based test platforms, hardware-in-the-loop automation, and AI-driven evaluation pipelines that hold complex systems, from XR devices to aerospace, to a measurable standard. Twelve-plus years turning subjective judgment calls into data other engineers can trust.

Featured work → Windrow ↗ CrowdDodge ↗ eric@brah.ms
1,000s
Daily automated AI-assistant evals, up from dozens run by hand
400
User device deployment led ahead of Magic Leap One's commercial launch
$300K
Saved in automation costs by unifying three toolchains
Hrs → Min
Root-cause investigation time, via an AI-driven analysis platform
Side projects

What I build on my own time

A governance layer, a deep dive into agent guardrails, that gives observability into, and decides what, AI agents are allowed to do, and a crowd-forecasting tool that doubles as my sandbox for AI-assisted development.

github.com/ejbrahms/Windrow

Windrow

A governance layer for AI agents. Windrow sits between agents and the MCP tools they call, so every capability is granted on purpose and every call leaves a record: who is allowed to use the Gmail tools, who actually used them last week, and what has been quietly denied.

Live demo ↗ View on GitHub ↗
Usage dashboard ● Enforced live
Windrow usage dashboard showing the per-phase latency breakdown, calls-over-time chart, and recent-calls table with per-call outcomes and latency
01

Know who can do what

Ask which agents are allowed near your inbox, your files, or your calendar, and get a straight answer instead of grepping config files and hoping you found them all.

02

Nothing runs unasked

An agent reaching for a tool nobody gave it doesn't quietly get through. The call is stopped, or you're asked first, so new abilities don't arrive already switched on.

03

See what actually happened

Every call an agent made, and every one it was refused, is still there a week later, so a surprise can be traced back, and access nobody ended up needing can go.

Node.js Express SQLite React & Vite Model Context Protocol Agent Hooks
crowddodge.com

CrowdDodge

Combines machine-learning models, historical wait-time telemetry, and weather and seasonal signals to forecast crowd levels, recommend attraction routing, and estimate wait times across major theme parks.

Visit the site ↗
crowddodge.com ● Live
CrowdDodge trip planner showing the quietest days for a Walt Disney World visit
01

Forecasting models

ML models trained on historical throughput data to predict crowds and wait times before you arrive.

02

An AI tooling sandbox

Where I prototype with new LLMs, agentic workflows, and interface design outside of work constraints.

03

Itinerary optimizer

Plans a day route algorithmically to cut queue time and fit more into the day.

Python Machine Learning AI Dev Tools Time-Series Analytics REST & Web APIs
Career

Where this was built

Twelve-plus years across two industries where "it works" has to be measurable: XR spatial computing, and commercial and military jet engine programs.

2024–Present

Magic Leap

Software Engineering Director (Technical Lead)
System Analysis, Verification, and Validation (SAVV) · Google XR Partnership
  • Built the engineering performance platform that became the source of truth for AI-assistant performance on prototype smart glasses, scaling from dozens of manually run scenarios to thousands of automated evaluations a day.
  • Designed a configuration-driven framework that scaled automated AI-assistant evaluation across 10+ tool integrations, including YouTube Music, Calendar, and Gmail.
  • Led hardware-in-the-loop infrastructure for prototype smart glasses, coordinating mechanical, software, and systems engineering into one automated validation pipeline.
  • Built the CLI tooling and integration layer connecting an Android XR device farm to CI/CD, with unattended 24/7 recovery.
  • Built an AI-driven analysis platform that automates root-cause investigation of hardware and software failures, cutting investigation time from hours to minutes.
2019–2024

Magic Leap

Software Engineering Director (Technical Lead), Senior Engineering Manager, Engineering Manager
System Analysis, Verification, and Validation (SAVV)
  • Led the engineering team that built a robotics-based measurement platform for XR spatial-computing performance: motion capture, display capture, computer vision, precision synchronization.
  • Drove adoption of that platform across perception, graphics, mechanical, electrical, and calibration teams.
  • Replaced subjective manual review with automated regression measurement for AR world-lock stability, quantifying registration error, drift, jitter, and latency.
  • Established hand-tracking benchmarks used to compare motion-to-photon latency, noise, and tracking error across competitive XR devices.
Jul 2018 – Dec 2019

Magic Leap

Systems Test & Integration Engineer (Contract)
  • Led technical execution of a 400-user device deployment ahead of Magic Leap One's commercial launch: OS provisioning, app setup, and checkout workflows.
  • Built telemetry analysis tools and reporting workflows that turned field data into commercial launch-readiness decisions.
Feb 2014 – Jul 2018

Belcan Corporation

Senior Software Engineer & Software Engineer
  • Built a rapid-prototyping and automation framework unifying C#, MATLAB, and Python into one execution environment, saving $300K in automation costs.
  • Designed telemetry analytics systems for commercial and military engine programs, turning flight-test and fleet sensor data into troubleshooting pipelines.
  • Built operational analytics engines processing ACARS/ACMF telemetry for fleet anomaly detection and FAA-required launch-readiness decisions.
Toolbox

What I build with

Platform architecture, robotics, spatial computing, AI workflows, and the software underneath all of it.

Languages

Python C++ C# SQL MATLAB

Infrastructure & Platform

Hardware-in-the-Loop Device Farm Architecture Git & Jujutsu Bazel Docker gRPC & Protobuf CI/CD Pipelines Embedded Systems

Automation & Tooling

AI-Assisted Analysis Workflow Automation Test Orchestration Validation Frameworks Android Devices CLI Tooling

Robotics, Vision & Data

OpenCV Sensor Fusion Camera Calibration Hand-Eye Calibration SLAM Validation Telemetry Pipelines Pandas & NumPy
Background

Education

Master of Science in Computer Science

Florida Atlantic University
Graduated December 2017

Bachelor of Science in Aerospace Engineering

University of Central Florida
Graduated May 2013