Independent research.
Practical questions.
I’m Jeff Dran. Drantech Labs is where I investigate ideas in computing, mathematics, and engineering, and test whether they can become something useful.
My interests span artificial intelligence, compression, symbolic computation, cryptography, and the behavior of complex systems. I’m drawn to the connections between these fields: how information is represented, how systems learn and adapt, and where a different approach might make a meaningful difference.
The current research directions include private AI, efficient computation, dependable automation, and ways to make technical results easier to verify. Physical engineering and materials science are longer-term interests, with their own requirements for measurement, equipment, and collaboration.
Start with what is known.
Before treating an idea as new, I want to understand the existing work. That means reading the literature, checking recent results, and choosing a useful baseline. From there, the aim is to build a small experiment that can answer a specific question—including when the answer is that the idea does not work.
Keep the evidence with the work.
A result is more useful when someone can understand how it was reached. I want the methods, inputs, versions, assumptions, and limitations to travel with anything I publish. AI is part of the toolkit for exploring ideas and building experiments; its output still needs to be checked against evidence.
Build toward something useful.
I’m interested in improvements that matter in practice: using less memory or energy, keeping data private, making systems more reliable, or solving a problem that existing tools handle poorly. Some investigations may lead to software, services, or physical products. Others may produce a useful explanation or show where to stop.
Share deliberately.
Public notes explain the thinking behind the work and any results ready to be shared. Early experiments and sensitive details stay private. Collaboration and access are considered project by project, with appropriate agreements where needed.