Leandro Seixas

Physicist · Professor · Full Stack Developer · AI Engineer

Computational materials scientist and full stack developer. Researching materials through computational simulations and machine learning, and developing the tools for these simulations.

See my work Academic CV

Based at Ilum – School of Science, CNPEM

Focus 2D materials & energy systems

Published 30 peer-reviewed papers

Things I’m building

Software that puts computational materials science in more hands.

Poraquê

Machine learning density-based operators.

Predicts the charge density and the kinetic energy density of a crystal straight from its geometry, skipping the self-consistency cycle that makes DFT expensive.

PythonIn development

poraque.seixas.dev Source

Calango Studio

Visual atomistic modeling.

Builds, edits and inspects atomic structures visually, with a C++ core that stays responsive on systems large enough to stall a typical viewer.

C++In development

calango-studio.seixas.dev Source

Mandacaru

Fermionic quantum simulation, one API.

Solves the fermionic Hamiltonian of a molecule or crystal with VQE and ADAPT-VQE, running the same script on a simulator or on real IBM, Amazon Braket and Google quantum hardware.

PythonIn development

mandacaru.seixas.dev Source

All projects

Four dimensions, one practice

The same questions keep showing up in different rooms. These are the rooms.

Physicist

Electronic structure of low-dimensional materials, from topological insulators to phosphorene, studied from first principles.

DFT2D MaterialsQuantum MaterialsAlloysAb Initio Simulations

Professor

Nine years of higher education teaching across quantum mechanics, thermodynamics, solid state physics and programming.

Solid state physicsQuantum mechanicsThermodynamicsProgramming

Scientific Applications Full Stack Developer

Frontend and Backend

Research software carried the whole way: the numerical core and the API that serves it, the interface that makes a result readable, and the deployment that keeps it available to the people who need it.

PythonC++FlaskHPCOpen source

AI Engineer

Models for the physical sciences

Models that learn physics from simulation data and then stand in for it: neural operators that predict electron densities directly from a crystal geometry, and machine-learned interatomic potentials that run the dynamics ab initio methods cannot reach. The whole pipeline is the job, from generating and curating the training set to validating a model against the physics and serving it in production.

PyTorchNeural operatorsML interatomic potentialsTraining pipelinesMaterials informatics

A short introduction

I am a computational materials scientist specialising in ab initio simulations and machine learning for atomistic modelling. My research centres on two-dimensional materials and energy systems, combining Density Functional Theory with data-driven methods for predictive materials design.

That path started in theoretical physics, in topological insulators, phosphorene and other two-dimensional materials, and moved steadily towards the computational side: first the simulations, then the software that runs them, and now the models that let us skip ahead of the calculation entirely.

Somewhere along the way the software stopped being a side effect of the research and became part of the work itself. I build scientific applications full stack: the numerical core and the API that serves it on the backend, the interface that makes a result readable on the frontend, and the deployment that keeps the whole thing running.

Over ten years of research and nine years of university teaching sit behind that, alongside more than thirty papers and the tools I build to make this kind of work easier to reach.

Get in touch

Selected research

Spontaneous chemical functionalization via coordination of Au single atoms on monolayer MoS₂

Sci. Adv. 6, eabc9308 (2020)

Multiferroic Two-Dimensional Materials

Phys. Rev. Lett. 116, 206803 (2016)

Vertical twinning of the Dirac cone at the interface between topological insulators and semiconductors

Nat. Commun. 6, 7630 (2015)

All publications