Although you can also find my articles on Google Scholar or ResearchGate, you will find them here in a timeline and with brief comments. Besides the given excerpts, click on the title for more elaborations.
21 publications grouped by career stage:
In this ICLR 2026 work, we revisited diffusion forecasting from first principles: by aligning diffusion time with physical time and estimating schedules from data, TEDM becomes both faster and more accurate on long-horizon benchmarks.
A tracking framework that models uncertainty in detections for more robust visual multi-object tracking.
Ideas from non-equilibrium thermodynamics coming back! In the GAN game for supervised image-to-image translation, the generator and discriminator don’t share information symmetrically, usually hindering equilibration … Maxwell’s demon?
Transferring style in unsupervised image-to-image translation was first cleverly done using cycle-consistent GANs. I noticed that similar generative modeling ideas could be used to let an artificial agent learn trajectories (geometry + evolution) by imitation.
I was asked if I wanted to supervise a Master student interested in applications of reinforcement learning. Why not? I was eager to learn! Fortunately, he was very motivated and ended up doing a nice work for predicting vessels trajectories.
Entering the field of anomaly detection using AI, I lead a team to develop a simple mechanism to detect abnormal vessel motion from video and notify it via RESTful services to a UI indicating the corresponding areas on a map.
Shortly after beginning my new job as an AI researcher, I got interested in the field of uncertainty estimation in ML models. I noticed that estimating confidence intervals was similar to computing the chemical potential of an electron gas. I then began the journey.
Quantum magnetism comes back! I discovered an old problem related to a linear contribution to the resistivity of ferromagnets around liquid helium temperatures. I showed that this could be due to electrons interacting with chiral phonons.
How are thermodynamic phenomena to be interpreted at quantum scales? Failure to do this properly might lead to paradoxes, as in the Jaynes-Cummings discussion of the one-atom maser. This was a suitable testbed for my entropy-production theory.
In quantum mechanics, we talk about quantum observables as the entities that are measurable. In termodynamics, the entropy is a state variable, as measurable as the temperature and pressure. What is the corresponding quantum observable?
I was captivated by Fano’s formulation of the density matrix as a vector in Liouville space. Using it for the quantum particle in a box problem, I arrived at a novel description of its back-and-forth motion and used it to interpret the quantum Newton’s cradle.
At some point during my PhD, I couldn’t reconcile in my mind having isolated quantum systems which, after some initial manipulation, come to equilibrium after a long time. So I looked for sufficient conditions for this to happen, arriving at a very restricted class.
I had studied enough ground-state and finite-temperature properties in spin chains for my thesis. Nevertheless, I felt that something was missing: how the behavior of initial states evolves with time. So I closed the chapter by investigating this for a special case.
How much can the quantum entanglement within a mixed-spin chain survive when the temperature is increased? In this study, I could figure out means to affect that threshold temperature by varying the crystal-field anisotropy.
It is very satisfying when a material has an unexpected behavior and you can find a model which explains it. By swimming in the sea of interesting problems, I found this one … and had a lot of fun realizing what was going on.
Entanglement is a weird quantum phenomenon and its relationship with quantum phase transitions is fascinating. Evidently, exploring this connection within the context of mixed-spin systems was the next step.
What I had observed in the (1,3/2) spin chains was too beautiful to leave it unfinished. I couldn’t resist generalizing the methods that I had used to other spin configurations. This time, however, the soup had more ingredients.
Now it is time to complicate things and study more general cases of spin configurations and interactions. I wanted to lift a controversy regarding the effectiveness of the spin-wave theory when applied to mixed-spin chains with crystal-field anisotropy.
The Master comes with new challenges … and I began to embrace them by fixing my attention on quantum magnetism in low dimensional systems. How else to understand the topic than by starting from simple specific examples?
I couldn’t let my Bachelor finish without answering how the corrected bending-moment formulation of cantilever beams that I had discovered could be applied to materials which don’t have a linear elastic response.
As a Bachelor student, I was confronted with studying the bending of cantilever beams in a experimental physics class. After realizing that the theory was incomplete, I exploited the consequences of a proper formulation.