About
As a PhD student in Bioinformatics I work on software engineering, machine learning, omics, structural biology and chemoinformatics. My research applies deep learning, focusing on generative models, to develop an antimicrobial language that generates synthetic compounds inspired by the immune systems of diverse organisms. The goal is actionable routes to new therapeutics against antimicrobial resistance using scalable in silico methods.
I am a Nextflow Ambassador, a specialist in data science and analytics and an MBA student in software engineering, which grounds the research work in reproducible workflows, data governance and software architecture.
During my MSc in Genetics and Molecular Biology I developed AMPidentifier, a machine learning tool that identifies antimicrobial sequences in genomic and proteomic data. My BSc is in Biomedical Sciences, with a concentration in bioinformatics and clinical pathology.
Skills
6 groupsExperience
9 rolesEducation
5 degreesCertifications
13 itemsPublications
7 papersTalks
22 talksAwards
11 awardsTeaching
8 coursesMaterials
9 reposCode
6 reposBio Signal Hub
Bio Signal Hub is also available as a Python CLI providing a hub of tools for structural bioinformatics analyses. The website offers automated graph generation and 3D protein visualisation. Both are under active development.
PyMol Vitral
Visualisation presets for PyMOL, aimed at molecular simulation systems: lipid bilayers, proteins and peptides. The protocol splits the loaded system into independent objects, applies a calibrated material and exposes numbered presets that combine one representation per component. It runs on PyMOL 2.x, open-source or incentive, with no external dependencies.
Ghost in the Fold
Fourteen models read one by one: AlphaFold 2 and 3, the ESM line from ESM-2 to ESM C and ESMFold2, ProteinMPNN, LigandMPNN and the RFdiffusion family. Seven of them carry a clickable architecture diagram, component by component, with the tensor each one passes on. Every number traces to the paper or the model card it came from, and the weights column was checked against the registry. Seven interactive labs run the mechanics in the browser: masked prediction, a real attention head, and frozen against full against LoRA adaptation. English and Portuguese.
Gallery
6 images