About
I study and grow learning systems that can organize themselves.
I was born in San Diego in the middle of the 90s. I've always had an interest in science and philosophy, but I probably became interested in neuroscience and AI due to chronic childhood migraines and a concussion which led to (temporary) radical changes in my conscious experience. In addition to this interest in mind and consciousness, I've always loved mathematics and its application to furthering understanding of reality.
You will probably notice that I like loops: recurrent networks, attractors, the cancellation of forward and backward predictions, energy-based models, agents that distinguish self from world by a sensorimotor loop. A lot of my projects are just paper implementations to get a sense at what others are working on, though I often experiment beyond them or extend them. Upon learning about the backpropagation algorithm, I was struck by horror: this does not loop well and animal brains are all loops.
The name of this site, endopoietic, is partly a play on "autopoietic" but slightly inverting "self-creation" to "creation from within".
What I work on
- Learning without backprop: Hebbian rules, predictive coding, forward-only and evolutionary learning, benchmarked fairly.
- Memory & the hippocampus: Grid cells, attractor networks, and associative memory that learns online.
- Self-supervised world models: Joint-embedding predictive architectures, from images to agents that learn skills.
- Dynamical systems & time: Koopman operators, HiPPO memories, and forecasting chaotic systems.
- Generation & pattern formation: Audio codecs, self-organizing codebooks, diffusion, and growth models from biology.
- Mathematical foundations: Compression, fractal geometry and information geometry of learning systems.
Get in touch
Feel free to reach out via LinkedIn, Github, or by inferring a likely email on this site. I'm open to MLE and research engineering roles and collaborations in machine learning and computational neuroscience.