My Definition of “Full Stack Engineer” Is Changing Fast

I no longer believe a junior web developer is a viable career path. Senior web developer is next on the chopping block. I manage my AI agent like it's a junior dev. I only provide prompt engineering, then submit PR code reviews and change requests. The exact same way I used to work with a human junior. The AI faster, broader and tireless. It adds skills I used to have to subcontract, like graphic design. It can help me translate coding languages to work in a framework I've never used.

My code editor (Cursor) is surely storing my agent's output and my human feedback as a training set for reinforcement learning with human feedback (RLHF). As this happens across the world with millions of developers, my senior dev job security continues to erode.

So what's my solution? Expand the stack.

Its obviously a general term but to me “Full stack” used to mean across web and mobile devices: a developer who could handle frontend to backend and basic DevOps. That was enough.

But today the definition is expanding in both directions, up into AI, and down into hardware. The STEM professionals at every layer who ignore this shift risk falling behind.

Up the Stack: The AI Layer

AI is no longer a tool, it’s a layer. At Rose City Robotics, we don’t debate whether to use LLMs in the software development process we start with them.  A full-stack engineer on our team is not only expected to write code in an AI-integrated text editor, I expect them be able to fine-tune language models, and integrate AI interfaces.

But the real up-stack leap goes beyond prompt engineering. It’s when our developers start building image classifiers, tuning YOLO models, deploying vision transformer models on edge devices, and building custom machine learning models in PyTorch across language, vision, and sensor fusion. That’s the new software stack.

Down the Stack: Physical AI

At the other end of the stack has always been hardware and historically it's been the domain of mechanical and electrical engineers, rightfully so. 

The best example of this is the traditional automation industry. Think big industrial arms like FANUC, ABB and Kuka. Technology that has been used in processes like auto manufacturing since the 1960's. And the software stack feels like it. 

Nearly every arm manufacturer has their own proprietary hardware and software stack, like you have to learn their custom scripting language which reads like COBOL. But it's precise, repeatable and it solid, just like the banks still running similar architectures on ancient mainframes.

Welcome the new age of open source hardware and software stacks like the latest Robot Operating System (ROS2) and projects like Stanford's Mobile ALOHA open source hardware and software autonomous robot. The GitHub repository includes the training algorithms for a transformer neural network, providing a template for engineers to teach their own robot through demonstration instead of code, called imitation learning.

The giants are even starting to come around, FANUC just released an officially supported FANUC ROS2 Driver on GitHub. One of our students was there to brush up on ROS2 just for this, his shop is a FANUC shop and they are starting to integrate computer vision. ROS2 provides the manufacturing automation team with a framework for prototyping. That's the beauty of open source frameworks, you aren't reinventing the wheel but you can always fork the repository and tweak the gear ratios.

Mechanical and electrical engineering, mechatronics, embedded systems, machine learning and software are all merging. We need teams of these experts to cohesively share information and systems, they currently don't. 

For software engineers it means understanding how latency plays out on a robot and how device limitations like RAM, ROM and battery life impact your training data collection. What happens when you use this robot in real world conditions, not in a lab. What about on its 1,000th battery cycle?

It's Physical AI where software meets hardware and autonomy gets real world.

I introduce the Physical AI Full Stack Engineer, there aren't many out there at the Senior level, they are rare even at the Junior. It takes an incredibly diverse set of skills, cleverness and determination.

Meet Max Blanksby, a high school junior who just earned his certificate of completion in our second cohort of Engineer Boldly: AI Robotics Sprint.

Max is an AI-native student, with unabashed use and customization of LLMs. But the metal is there too. He quickly commanded and began customizing his TinyAI Stack, the autonomous-ready mobile base robot with vision, LiDAR, ROS2, and onboard AI inference capabilities we give to students in our hands on bootcamp. 

Linux and Python, he was already a pro. For ROS2, he just needed the docs and the basics and he was off on his own. Fine tuning vision models and training a transformer, yeah he's familiar but not down in the weeds of the math which he voraciously consumed, switching to his history homework while the rest of the class caught up.

Max's ambition? Quantum computing.

This story repeats, Kenny Ma was one of our first high school interns who helped us shape the SLAM curriculum using an off the shelf Turtlebot3 we got for him. Kenny learned Linux to build out functioning LiDAR based SLAM mapping with ROS2, in 6 weeks, mostly on his own, while he completed high school. He is now at NYU studying mechanical engineering.

Why It Matters

The engineers we train aren’t just writing code and running commands, or driving toy robots. They’re shipping perception systems, connecting motor drivers, collecting training data, and deploying real-time onboard inference to robots that roll across a lab floor autonomously.

At Rose City Robotics, We’re Redefining the Stack

The students in our first two cohorts include senior automation engineers, senior web developers and software engineers, embedded systems techs at hyper scalers, mechatronics engineers and of course the most formidable, the high-school hackers. Over three quarters have advanced degrees, several with decades of experience. They all recognize the time for robotics is now.

Our cohorts build AI-ready robots, publish GitHub repos, and learn from a PhD who has shipped missile trajectory algorithms at Northrop Grumman and written the code on medical devices used inside human patients at YorLabs.

Engineer Boldly. Learn Deeply. Lead the Future.

Ready to upgrade your stack? Join the next AI Robotics Sprint and build something that moves. This is the program for elite engineers across a variety of advanced STEM backgrounds looking to grow their definition of full stack towards the edge.