← home

Antonio Llano

I'm an Electrical Engineering and Computer Science student at Stanford University in California, USA, where I further my interests on alternative computing architectures and embodied ML. Before, I've worked on DNA Computing, Trading Algorithms, Film Production Software, and startups [1] [2]. I did computing and mathematics olympiads in HS, placing 1st nationally in the former, and participating in the IOI & IMO training camps.

Email  /  CV  /  Bio  /  X  /  Github

profile photo

Current Research Interests

I'm working on two threads: (1) RL agents that learn by prompting, or dreaming, in their own lightweight world models to improve sample efficiency in simulators or real-world experience; and (2) DNA-based molecular computers that sense, compute, and actuate for personalized drug delivery using Bayesian inference in chemical reaction networks.

Active Projects

Dream Learning in World Models

Developing open-source 2D environments where reinforcement learning agents generate and learn from self-prompted simulated experiences within learned world models, improving sample efficiency in real environments.

Molecular Computing for Targeted Therapeutics

Integrating Bayesian inference in Chemical Reaction Networks to create adaptive drug delivery systems capable of patient-specific sensing, computation, and therapeutic actuation at the molecular scale.

Timeline

Film production software at Sony Pictures Entertainment at 15. Worked at a couple agrotech startups in between stints.
SONY
Film
DNA computing research with Prof. John Reif at Duke University
DUKE
Research
Quantitative trading research at Intelneuron's Fujur Fund.
FUJUR
Trading
Stanford EECS studies (2024-present)
University
Cofounded a startup called W37.ai to build videogames with open worlds and narratives. Raised a small round from Accel at $5M valuation.
W37
Startup
Worked as an MLE at Luzid, the Prod company eating enterprise software implementation.
LUZID
Startup

Recent 10hr hack

Livestream Diffusion Exercise Bike with Custom Hardware Rig From Scratch in 10 Hours

On Raspberry Pi 4, low-res mesh with livestream diffusion, and a voice coach.

AI Bike ($9k OpenAI/Inworld prize)

Other Questions I'm Interested In

  • Could a TRIBE-like multimodal brain encoder be trained on motor task execution data to generate synthetic neural signals for use as reward functions in robot reinforcement learning?