About

I’m Marshall. I work at NASA Goddard as an AI Engineer. Before that I was at JPL. I have an MS and BS in Mechanical Engineering. I build and train large language models, develop AI tools for engineers, and work on metal additive manufacturing. I built rockets, robots, viruses, video games, guitars, turrets, mesh networks, terminals — literally anything I could think of and not get sent to prison for doing. Often these things were at odds with that final guideline.

This site is sort of a living memorial. It’s a place I can publicly document my work for an imaginary audience. Will some people see it? Sure. But it’s more important to me that there’s some record online of my living and life in general. It feels like I’ve done too much not to allow that.

What I Do Now

At Goddard I helped develop ChatGSFC, an internal AI system for NASA engineers. I led the development of its image generation capabilities and now work on LLM training and inference more broadly. I initially worked on an early text-to-CAD system before moving into more general machine learning engineering — both training and inference of these systems. Largely, that’s still what I do today.

My long-term goal is to help NASA become a leading government entity for artificial intelligence. If I can help thousands of engineers be more productive using the tools and models I develop, that’s far more resourceful than being one engineer working unaided.

I also built Ryngo, an AI-native terminal emulator. Beginners using Claude Code or Codex can ship faster and at lower cost. Power users can manage more agents and interact with their systems at a much deeper level than previously imagined. It’s the kind of thing that only exists because I got tired of the terminal not knowing what an AI agent was.

I’m also currently working on metal additive manufacturing, and writing a technical handbook on large language models aimed at engineers who want to understand the internal math, architecture, and implementation rather than just the API calls.

Research

My graduate research was through the ARCS Fellowship at NASA JPL, in collaboration with Honeywell. I worked on additively manufactured heat exchangers using TPMS (triply periodic minimal surface) structures with Dr. Bingbing Li and Dr. Changyu Ma. We explored how complex lattice geometries in aluminum and Haynes 282 can dramatically improve thermal performance. That work resulted in published research:

• TPMS Heat Exchanger Research — MDPI Biomimetics
• Additive Manufacturing Research — IJES

The Early Years

I have been building things for as long as I can remember. I was always the builder and the fixer. By twelve I was the resident mechanic in my friend group — fixing airsoft guns, designing new components, drawing up intricate engineering diagrams of firearm mechanisms in my bookbag. Having those diagrams in your bookbag does create problems, as it turns out.

In fifth grade I built a maglev train demonstration model. A 1/24 scale car kit riding on a meter-long wooden track with counterbored holes and seated internal magnets — essentially a way to showcase magnetic resistance. It was the kind of project where your teacher looks at it and isn’t entirely sure you built it yourself.

Then Lego Mindstorms and robotics teams from about age ten to fifteen. That ended when I got asked to leave for writing a worm-style virus in BASIC and distributing it to peers in the after-school course. Which of course then led to all these interests combining in increasingly ambitious ways.

The Turret

At seventeen I built a fully automated turret and strapped it to the roof of my car.

It connected to my laptop and had its own power supply with a couple NEMA 23 stepper motors — one for pitch, one for yaw — mounted on a lazy susan bearing with two round MDF blanks as the rotating base. The laptop ran OpenCV for object detection and was wired to an electric airsoft gun which had a webcam attached. The webcam streamed to the laptop, which would position the camera at the center of the bounding box of whatever object you wanted, using the stepper motors, and then using a relay connected to the gun’s MOSFET it would fire.

Spray painted black, of course.

I didn’t always make weapons and things of poor nature. I spent three years in my high school’s woodshop where I made bowls, cabinets, and helped around the school voluntarily. I was told many times that repeating a course looked bad on college applications. I currently hold multiple engineering degrees and have never had an interviewer question my love of woodshop.

College

I joined the CSUN Formula SAE team where I volunteered as team lead for both the EE and ICE powertrain teams. Even though I wasn’t a senior and therefore couldn’t technically participate since it was senior design, my name was added to the car’s wings as a team sponsor.

I took a break from projects during the pandemic and instead focused on maximizing the number of courses I took. Which then gave me the time to have fellowships and research opportunities for the latter half of my undergrad. That gave me the time and opportunity to work on the NASA JPL CADRE missions and JPL PUFFER during my senior design, which also allowed me to continue working on the heat exchangers and additive manufacturing. That would then culminate in moving to NASA Goddard.

The Hackathon

I built a mesh networking system that won a hackathon. The work was subsequently deployed by the Ukrainian Ministry of Defense and other organizations operating in contested environments. One of those projects where you build something over a weekend and then get a phone call a few months later that changes the context of what you made entirely.

The Guitar

The project I’m most proud of is probably the guitar.

My dad named me Marshall because of his work with James (Jim) Marshall — the founder of Marshall Amplification. He wrote four, maybe five books on the topic. He helped run Guitar Center for over twenty years and worked at Fender in the nineties. I intentionally removed myself from that industry and that world for a long time.

But eventually a small topology optimization project turned into engineering a truly novel instrument — a Voronoi lattice Stratocaster body, 3D printed in resin, electroplated in nickel, assembled by the Fender Custom Shop, and showcased at NAMM. Going from not playing guitar to allowing a small project to make a truly novel thing for the world and for the two of us is great. My dad could have or play any guitar he would ever want, but he’s always most proud of the guitar that I engineered.

So that’s cool. A close second would be the work I’m not allowed to talk about.

How I Work

I like to work in 40-hour sprints with a 4-hour rest period in the middle. Doesn’t matter if it’s deep research or rapid prototyping — I’ve done and love both. I burn out from not seeing progress, not from overworking myself.

I’d also like to build a company at some point. I believe that having equity in your work is a very important and underrated concept in the modern economy.

What I Think

I think AI is going to change white-collar work the most, but not in the way most people think. What AI has really shown is that self-driven people who don’t have to be asked to do something are going to become the greatest assets for organizations and society. The age of white-collar work where one waits for a task seems to be ending — but at least the people performing these tasks have an always-on omnipotent super intelligence to help them along the way.

I think that in twenty years we’ll see a massive amount of manufacturing reshored in America, with lights-off factories producing higher-quality, lower-cost goods for a larger population. Not saying I agree with that as a romantic ideal — the idea of running a Series 1 Bridgeport and driving a V8 station wagon will always be the way I see the manufacturing job as it could and should be. But that’s a romanticization of the actual work at hand, not the reality.

The most powerful technologies emerge when theoretical knowledge and physical engineering meet. That’s where I try to operate.