Worked since the inception of a project to build a comprehensive AI platform focusing on shareable rateable user-generated AI artifacts (e.g.: threads, prompts, chatbots, datasources for RAG, plugins).After almost two years, the platform supported over 15 apps, including a VSCode extension with a Copilot-like feature, a Chrome extension with web interaction plugins, a canvas-based webpage for creating and iterating on artifacts, files, and images, and an SPA for documenting software projects using LLMs.As the only security-conscious dev of a small team (15 members), I dedicated a portion of my time to harden our applications:- implementing an authentication middleware from scratch to be bundled on all microservices (https://github.com/TeiaLabs/tauth)- writing specs for a mongodb-backed simple authorization microservice.- demanding 2FA on Atlas, AWS, Zilliz, and other mission critical third-party services.- safeguarding access to prod data from everyone.- hosting architecture review meetings for new services.- pushing colleagues to use password managers to store their database credentials and API keys.- enforcing least privilege wherever possible.

Proprietary research projects for OSF Digital.

Application security contractor for Grubhub.Web3 consultant for Willowtree.

Data warehouse project for HPe.

Designed, performed, and evaluated experiments with different neural network architectures inPyTorch to use a single U-Net with multiple decoders to denoise digital images (aka seeing in the dark,night vision)Refactored an object relationship project (visual relationship detection) to create amulti-dataset benchmarking CLI app on top of it. This was research work sponsored by Motorola to improve mobile cameras.

Deep learning research for big Oil & Gas companies.Auxiliei na elaboração da arquitetura e no desenvolvimento de pipelines baseadas em deep learning para uso em determinadas fases da exploração de petróleo para a empresa Shell.Várias pequenas PoCs e experimentos com deep learning usando PyTorch.Filtragem e adaptação de um conjunto de dados de imagens provido pelo cliente; treino de classificadores nos dados deles.Modelagem conceitual e implementação de um banco de dados utilizando SQLAlchemy para um projeto interno.

Agrobusiness meets Ethereum smart contracts."Innovative Business Solutions Developer"Participei do desenvolvimento de uma Proof of Concept que tornou digital um processo de negócio de um sistema de gerenciamento de contratos agrícolas da SAP integrando-o a blockchain pública Ethereum.Posteriormente participei do desenvolvimento de uma plataforma web (um DApp) utilizando a blockchain em conjunto com o SAP ACM como seus backends para as cooperativas agrícolas, os fazendeiros e os laboratórios que fazem análises da qualidade de grãos.

Feature engineering for a decision-tree-based machine learning algorithm. We used Fourier transforms to shift our data points to the frequency-domain so we could use frequency-related features as well as the several usual time-domain features. Libs used: Pandas, SciPy, MatPlotLib and Flask. Some Octave (MatLab) and a lot of reading up on C audio libraries (that's where we found the frequency features to be most efficiently implemented).I kept the codebase linted, type-checked and up-to-date with the latest PEPs (Python3.6).It was a research project, so agile (scrum+xp) was the way to go: Weekly meetings with the POs and a lot of pair-programming.

Answering questions about the ERPs;Reporting bugs to the development team;Updating the FAQ page;Initial requirements analysis for new tools, shortcuts and features for the main ERP.

(NET Telecomunicações outsourced its tech support to this company)Duties included:Monitoring NET’s (famous local ISP) hybrid fiber-coaxial network to report instabilities.Calling NET clients who were experiencing some sort of technical difficulty and helping them through it to avoid sending a repair crew unnecessarily.

Perfected the "pending requests" revision process which increased efficacy from ≈ 40% to 99%;Dealt with all overdue requests from ‘09, ‘10 and ’11;Color-coded urgent requests for quick reviews.

Created a spreadsheet to track arrival, repair and status of computers and peripherals;Assembled an octo-monitor setup for the new CCTV team;Rearranged the office furniture freeing previously inaccessible bench space for up to 33% more repairs.