Thiago Maffei Email & Phone Number

São Paulo, Sp, Br

o Responsible for the digital transformation process for RD Medicine, an EdTech company assisting Brazilian students and medical graduates in preparing for medical residency in the USA.o Acted as a Full-Stack Engineer, responsible for the implementation and structuring of the company's cloud-native infrastructure using AWS.o Implemented video streaming solutions using S3 buckets for storage and CloudFront for content delivery.o Managed credential authentication with AWS Cognito and user traffic with Elastic Load Balancing (ELB).o Utilized Python (Flask) for backend development, Next.js with Tailwind CSS for web frontend, React Native for mobile, and NoSQL (DynamoDB) and SQL (PostgreSQL) for database management.

Paris, Fr

o Conducted extensive literature reviews on data-driven methods to predict the State of Health and Remaining Useful Life of Li-ion batteries.o Utilized time series data to develop machine learning models for accurate predictions of battery indicators.o Implemented state-of-the-art models from scratch, including Transformers and BERT, to enhance predictive accuracy (<7% MAPE).o Developed a comprehensive data pipeline to process incoming data, perform necessary transformations, and train various machine learning models.o Benchmarked multiple models to identify the most effective approaches for predicting battery health and lifespan.o Applied Transfer Learning techniques to leverage models trained on one dataset for improved performance on another dataset.

Boston, Ma, Us

o Developed the first general 3D deep learning ultrasound solver for real-time tFUS navigation in individualized skull models.o Implemented various deep learning architectures, including U-Net, ResNet, and Generative Networks using PyTorch.o Utilized gradient accumulation to address GPU memory constraints during large batch processing (3D data from CT-scan and wave propagation simulations). Commonly used to fine-tune LLM models (large-scale optimization).o Achieved a final model with 1% MAPE at 200 kHz, performing forward passes in 3ms, significantly faster than mSOUND (Hybrid Angular Spectrum solver - 30s) simulations.o Implemented a Holographic Reconstruction method for 3D beam field analysis.o Presented work at the 3rd Brain and Human Body Modeling Conference in Boston; currently co-authoring a peer-reviewed paper.

Paris, Fr

o Optimization of experimental photoreaction conditions using Bayesian Optimization.o Attributions: implemented Bayesian Neural Networks (BNN), Machine learning regression models, transfer learning and fine-tuning.

Paris, Ile-De-France, Fr

o Development of a system for automating mechanical testing of lipsticks for a leading French beauty company.o Attributions: development of the experimental system with Arduino; analysis of the results; presentation of the results to R&D employees of the company.

Paris, Île-De-France, Fr

o Study of different neural network architectures (DNN, ResNet) and exploration of Neural ODEs applied to dynamic / chaotic systems. Implementation of adaptive step size using Euler 2-step method in stochastic gradient descent (SGD).o Attributions: carrying out bibliographic research; development of Python code with PyTorch.

São Paulo, Sp, Br

o BTG Pactual is the largest Investment Bank in Latin America with 37 years of history in the Financial Market.o Worked in support to Sales & Trading (mainly Energy and Fixed Income, Currency & Commodities), Credit and Digital Banking areas. o Python development using frameworks for data manipulation, front-end and back-end.o Web applications/systems - Flask, JavaScript, Bootstrap, SQL and VBA.

São Paulo, Sp, Br

o SMaRT is a unique research group in Brazil that operates in the area of characterization and synthesis of composite materials, more specifically soft actuators.o Developed activities ranging from synthesis of composites to electrical, thermal, electromechanical and viscoelastic characterizations of IPMCs. o Built experimental apparatus such as systems for humidity and temperature control with Arduino and systems for electromechanical tests also with Arduino, symmetrical source and current amplifier.

São Paulo, São Paulo, Br

o São Paulo Research Foundation is a public foundation, funded by the taxpayer in the state of São Paulo, with the mission to support research projects in all fields of knowledge. In addition, FAPESP is known for its rigorous selection process for new fellows.o Developed a project related to the study of water absorption kinetics in Ionomeric Polymer-Metal Composites (IPMCs), as well as the influence of counterions and their mixtures in this kinetic phenomenon.o Prepared two scientific reports, in addition to writing an article in the synthesis area of IPMCs that is being reviewed by members of the research group.

Master Thesis, Computational Physics

Master Of Science - Ms, Computer Science

Bachelor'S Degree, Materials Engineering