Rodolfo Capdevilla, Ph.D.

Grandma is my hero. There is no braver woman in the world. She taught Amazons how to fight, and Valkyries how to ride Pegasus. She taught me the value of hard work, by doing it, not talking about it.

I love physics, research, and unveiling the mysteries of the Universe. The expansion of consciousness, the reach for the stars, the pursuit of knowledge drives me every morning.

Education Experience

Education

BS in Physics
BS in Physics
MS in Theoretical Physics
MS in Theoretical Physics
Ph.D. in Particle Physics
Ph.D. in Particle Physics

Experience

Postdoctoral Fellow
Postdoctoral Fellow
Research Associate
Research Associate
Faculty Lead, MS in Artificial Intelligence
Faculty Lead, MS in Artificial Intelligence

Teaching & Research

Teaching

Current Courses:

Deep Learning

Deep Learning

  • Deep Neural Networks
  • Convolutional Neural Networks
  • Autoencoders
Physical Properties of Matter

Physical Properties of Matter

  • Thermal properties of matter
  • Electromagnetic properties of matter (artifact)
  • Quantum properties of matter

Previous Courses:

Computer Vision

Computer Vision

  • Classical CV techniques
  • Convolutional Neural Networks
  • YOLO and Faster RCNN

Research

Neutrino–Dark Matter Interaction:

Majorana neutrinos convert into antineutrinos in a background of ultralight vector dark matter. This effect is suppressed by the small neutrino mass, but the enhancement by long astrophysical baselines can enable future searches for solar and supernova neutrinos to explore uncharted parameter space. The observation of a supernova neutrino burst at DUNE, Hyper-Kamiokande, and JUNO could probe dark matter masses beyond the capability of other future probes.

Gravitational waves:

A vertically polarized pulse (satellite 1) interacts with a GWB and some photons in the pulse transition to circular polarization. Half of the photons that transitioned pass through a birefringent material (satellite 2) as horizontally polarized and are read as signal photons by a single photon detector. If possible, the recycled pulse could bounce between satellites and increase the number of signal photons.

Radioactivity to electricity power conversion

Provisional Patent Application Number 63/874,784:

When radiation from the sample penetrates a material, a large population of ions form. Our simulation is a Monte Carlo radiation-transport engine that tracks individual decay particles (alpha, beta, or gamma) emitted from a radioisotope source as they propagate through a host medium, sampling discrete interaction sites and the resulting electromagnetic shower of secondaries.

Charged-particle slowing-down is modeled including multiple-Coulomb scattering, while photon transport uses mass-attenuation coefficients with Klein-Nishina sampling of the interaction channels. Local energy deposition along each track is converted to ion-pair yield through the material's mean ionization energy.

Reactor vacuum chamber
Ion Density (pairs / cm^3 / s)

ODY (Object Detection for You)

ODY logo

ODY is a desktop application that overlays a transparent, red-bordered capture window directly on the screen, continuously grabbing whatever appears inside it every second and running it through an object detection model. Users define up to three target object labels as chips (e.g., person, car, dog) and whenever the model detects a match within the captured region, the app plays a configurable alert sound and saves the annotated frame to a rolling local image buffer. The interface is usable for both real-time surveillance scenarios and one-off visual checks without requiring any external server or internet connection.

Simple demo for visualization. The application will be launched soon.

Weather Anomaly Detection Autoencoders

WeatherAE logo

WeatherAE is an unsupervised rain-nowcasting system for Miami. For a chosen weather station and reference day, it pools ten years of hourly observations from that station within a +/-20-day seasonal window, then trains 24 hour-of-day autoencoders exclusively on dry hours. The whole pipeline runs as a Streamlit app with three sections: Data Visualization, Model Training, and Results Records. Data retrieved from Open-Meteo Historical Weather API.

Dry training samples vs hour of day
Reconstruction error vs hour of day
Stations used map

Workflow Automations using AI

As Faculty Lead of a MS program, I need to perform a series of operational tasks that can be repetitive. This includes checking the status of the courses; assignments, attendance, the health and status of classes. Before arriving to Atlantis University these tasks were done manually. I have created agents that automate these processes, and free faculty leads time for research of other activities.

Version for visualization. The data is retrieved without APIs.
Version for visualization. The data is retrieved without APIs.