In a recent adventure, I embedded Frigate into my custom DAX home automation system by placing an nginx reverse proxy in front of DAX and routing Frigate through a /frigate/ subdirectory. This solved same-origin and streaming issues while preserving HTTPS and full WebSocket/MSE functionality. The final result is a seamless, reliable, real-time camera interface inside DAX without modifying Frigate or the DAX python core. Only a few changes on the DAX web interface for embedding the iframe were required.

It’s been just over two years since I joined Hermeus as a Principal Avionics Engineer in November 2023, and it has been an absolutely incredible journey. For those unfamiliar, Hermeus is a defense-tech startup with a singular focus: the rapid design, build, and test of high-Mach and hypersonic aircraft. It's been quite a whirlwind for myself, moving from working on small, low-powered IoT devices to fast, experimental airplanes.
The core engineering philosophy at Hermeus is iteration. We are building progressively more complex aircraft to solve problems in hardware, not just in simulation. Before I started, the team had just completed the “Mk 0,” a non-flying, airframe-shaped ground test vehicle. It was a critical step for validating core subsystems—a J85 engine, full hydraulics, HV power, avionics, and comms—and exercising our ground testing operations and capabilities.
I came on board just as the team was pivoting to Mk 1, the aircraft destined to be Hermeus’s first to fly.
The goal for Mk 1 was straightforward but immense: prove we could design, build, and fly a jet-powered, remotely-piloted aircraft, FAST. As a Principal Avionics Engineer, my work was split between foundational electrical system design and a critical, time-sensitive triage of the RF systems.
As my first assignment, I was responsible for designing several key electrical subsystems from scratch.
As I was wrapping up the HV system integration testing, our RF lead departed the company. With my background in RF systems and flight testing from GTRI, I volunteered to take over the comms systems.
This was a critical moment. The vehicle was already in field testing, and the lack of a reliable command and control (C2) and telemetry (TM) links was delaying the entire program.
The vehicle had two redundant, independent radio systems, and both were failing.
Over the next few months, I systematically diagnosed the issues and resolved our communications problems.
We also augmented the architecture, adding a SATCOM radio for beyond-line-of-sight (BLOS) capability and an additional line-of-sight (LOS) radio for redundancy.
One of the biggest process improvements I pushed for was adding better metrics to our software. Our telemetry monitoring was “blind” to RF performance, which made debugging performance issues very difficult. I worked with the software team to add real-time tracking of RSSI, packet reception rates, and other key metrics that were provided by the radios. This allowed us to stop guessing and start debugging with data.
With these new metrics, we ran dozens of tests. I wrote my own Python scripts to post-process and visualize the data. This analysis was key to building confidence and, eventually, getting all radio systems working reliably through low-speed and high-speed taxi testing.
On May 20, 2025, Mk 1 successfully completed its first flight. This was a monumental achievement for the company and a huge personal milestone. I'm incredibly proud to report that all the systems I was responsible for—the HV and LV power systems, the RIDD converter, and the entire C2/telemetry link—performed flawlessly.
We barely paused to celebrate. Work was already well underway on Mk 2.1, our supersonic aircraft slated to fly in early 2026. Mk 2.1 is a much larger aircraft, equipped with a Pratt & Whitney F100 engine.
I am continuing my role as RF Lead, but my responsibilities also expanded.
The entire process for Mk 2.1 has been a marked improvement from the scramble on Mk 1. Our methods, documentation, and integrated testing processes are substantially more mature. It’s the iterative development philosophy in action.
The pace at Hermeus is relentless, but the mission is clear and the engineering is deeply rewarding. We’re moving fast, and I can't wait to see Mk 2.1 take to the skies.
While working for MapLarge, we started investigating wireless SoC's that could provide the lowest possible power consumption. I found Nordic Semiconductor SoC's to be the best choice because of their ability to draw extremely low amounts of power in sleep, there are ample amounts of documentation, SDK examples and user examples out on the open web, and the parts have also been somewhat adopted by the open source communities. We built a variety of products with Bluetooth Low Energy, LTE-M, GPS and LoRa radios. We also spent considerable time learning and using Zephyr RTOS to obtain low power. Below are some of my favorite boards that I made for Nordic chips.
This Bluetooth Beacon was the simplest device we made, as it had only one SoC. The goal was to make a small device for basic sensor readings using BLE beacons, powered with a coin cell battery maximum battery life. At this point I have had several of these devices running since Oct 2021, so they have over 2 years of battery life. It's sleep current draw is 5 uA @ 3.0 V.
The LTE asset tracker as a much larger device that combined multiple radios into this IoT device. The main application ran on the nRF9160 LTE SoC, while the nRF52840 allowed for BLE connectivity to apps in addition to scanning for other BLE beacons to retrieve sensor data. There was also a Microchip WLR089U0 LoRa module for sending additional data over LoRa. This tracking device ran on either a rechargeable battery or 6 AA batteries for over a year of battery life. Its sleep current draw is < 100 uA @ 5V.
We had a proposal for a project that involved both BLE and LoRa, but it required some kind of display and no need for LTE. So, I designed another board that had only the nRF52840 and WLR089U0 and added an e-ink screen. This monochromatic screen only draws power (about 10mA) while it is updating the screen. This is the perfect screen for a device that only needs to occasionally update the screen. Unfortunately this project never made it off the ground so all I have is this prototype board. Maybe I will find another use for it some day.
Here are a few of my favorite embedded projects from work and personal projects over the years 2015 - 2019.
The ControlWidget was one of my first widely used generic microcontroller platforms. Based on a dsPIC, the platform was used to control various RF devices, front panel connections (buttons and status LED's), high voltage power switching, and temperature monitoring. It supported the following features:
The PicLeaf system was an evolution in how I architected digital control systems. Instead of using a fixed controller board, I created a multiple devices that all communicated over an RS-485 network. The PicLeaf Hub (pictured below) had an Ethernet port and two isolated RS-485 networks. Other PicLeaf devices could be connected to each network, and device addresses and type are configured by DIP switches. Software would run on a tablet computer, and a configuration file would inform the system which devices should be connected on the network. The software would then dynamically generate a GUI to control each device.
The PicLeaf Hub features the following:
Sometimes you need the processing power and network connectivity of a Pi but also an embedded device to take care of some faster tasks. This was the basis for my ESP32 gateway, which helped managed a variety of inputs into a data collection service running on the Raspberry Pi. My ESP32 Pi Gateway featured the following:
The Tea Earl Gray Hot Bot was my entry to the 2019 Alternative History Teacup Robot Race at DragonCon. This informal competition required a remote controlled robot with a tea theme to traverse an obstacle course. The robot features the following:
The Tea Earl Gray Hot Bot won first prize that year!
The beginning of 2023 marked ten years of me making printed circuit boards. From Eagle to Altium, through hole to QFN, one layer FR4 with no soldermask to 8 layers with exotic cores, it's been a fun ride! I hope to keep making cool stuff in the years to come.
Here are some blanks of my favorite PCB's that I've made and kept over the years.
I made my first PCB in 2012 when I was still in grad school. The PCB was designed to be a clone of an RFID card. Shortly after that I made a small adapter PCB to use a BusPirate to program the cards. You can see these in the bottom left of the picture.
My first “real design” (as I felt, at least) was the DoorProgrammer. The project was to allow access control for one of the old WREK studios. It was my first project where I designed the PCB, assembled the board, wrote and programmed the firmware, and wrote the software to communicate with it.