Publication
NLP for Sentiment Analysis: How It Works and Why It Matters for Business
https://www.toptal.com/developers/natural-language-processing/nlp-for-sentiment-analysis
Bruno is currently unavailable
Bruno Barbosa Miranda
Verified Expert in Engineering
Data Scientist and Python Developer
Belo Horizonte - State of Minas Gerais, Brazil
Toptal member since September 30, 2021
Bio
Bruno was originally a medical doctor but was the first to get a master's degree in computer science (CS) at one of the top universities in his country, without prior CS formal education. He is currently pursuing a Ph.D. in CS at the same university (UFMG). Bruno considers himself a highly motivated and passionate professional who loves his work and constantly learns new things.
Portfolio
PepsiCo
Data Science, Machine Learning, Python, Git, SQL, Pandas, NumPy, Spanish, SciPy...
Pixelcut Inc.
Machine Learning, Python, TypeScript, PyTorch, SQL...
Shop For A Better World
Azure Cloud Services, Cognitive Search, Classification, VM, LightGBM...
Experience
- Machine Learning - 10 years
- Python 3 - 8 years
- Scikit-learn - 6 years
- Deep Reinforcement Learning - 5 years
- Neural Networks - 5 years
- TensorFlow - 5 years
- Amazon Web Services (AWS) - 3 years
- LightGBM - 2 years
Work Experience
Senior Data Scientist (Machine Learning)
2025 - 2026
PepsiCo
- Handled machine learning (ML) initiatives for LATAM markets, supporting media investment decisions across multiple countries.
- Maintained Docker-based deployments and Streamlit dashboards, ensuring reliable and reproducible quarterly model cycles.
- Performed quarterly recalibration and drift analysis, improving model stability and forecast accuracy.
- Refactored and optimized Python pipelines to improve robustness and iteration speed.
Technologies: Data Science, Machine Learning, Python, Git, SQL, Pandas, NumPy, Spanish, SciPy, TensorFlow, Azure, Amazon Web Services (AWS)
Senior AI/ML Engineer
2023 - 2025
Pixelcut Inc.
- Deployed a computer vision neural network to estimate soft shadows based on cutout masks, with improvements on top of the reference paper.
- Implemented multi-GPU training with PyTorch-compiled code and multi-node disk access, training with over one billion synthetic image samples that I generated myself.
- Produced code for the deployment of the model into the client's app using FastAPI.
- Aided the development of a cutout masking model using synthetic data generation.
Technologies: Machine Learning, Python, TypeScript, PyTorch, SQL, Graphics Processing Unit (GPU), Google Cloud, Google Cloud ML, Google Cloud SQL, FastAPI, Deep Neural Networks (DNNs), Algorithms, 3D Images, Image Generation, NVIDIA CUDA, Models, Artificial General Intelligence (AGI), Generative Adversarial Networks (GANs), Communication, Large Language Models (LLMs), LangChain, LangGraph
Senior ML Engineer
2023 - 2023
Shop For A Better World
- Successfully classified a database of over 180,000 businesses in multiple categories to generate metadata.
- Configured Azure Cognitive Search for the client's search engine.
- Created a web scraping system that automatically updates database records over time while deduplicating instances obtained from different sources and consolidating the results.
Technologies: Azure Cloud Services, Cognitive Search, Classification, VM, LightGBM, Web Scraping, Deduplication, Database Management, Communication
Data Science Specialist Consultant
2023 - 2023
Faculdade Unimed
- Reduced the computation time of a critical software component from three days to less than five minutes.
- Helped build the entire Snowflake architecture from the ground up to host a software service for hospital stay analysis, similar to HRG and DRG.
- Built core functionalities to process the entire hospital stay analysis algorithm using Python and Snowpark.
Technologies: Snowflake, Docker, Python, SQL, NoSQL, Spark, Spark ML, Machine Learning Operations (MLOps), Sentiment Analysis, Regression, MySQL, Leadership, Communication, Large Language Models (LLMs)
Machine Learning Developer
2022 - 2023
Arthur Haliski De Andrade
- Deployed a genetic algorithm based on genetic programming that learns to generate custom technical variables for trading. The whole algorithm runs on GPU using RAPIDS.
- Delivered a mixed reinforcement learning neural network algorithm that learns to trade financial markets using neural networks written in PyTorch.
- Managed a team of three programmers while producing complex trading algorithms myself.
Technologies: Machine Learning, Python, Algorithms, Deep Neural Networks (DNNs), Genetic Algorithms, Trading Systems, Mathematics, Algorithmic Trading, Quantitative Research, Machine Learning Operations (MLOps), NVIDIA CUDA, Financial Modeling, Regression, Backtesting Trading Strategies, Models, Leadership, Communication
Senior Data Scientist
2022 - 2023
Microsoft
- Developed an email signature expansion method based on optical character recognition (OCR).
- Worked on clustering and classifying malicious emails.
- Resolved many new bugs on Microsoft services, tracking them with data science and other analysis tools.
Technologies: Python, C#, Azure, Data Science, Optical Character Recognition (OCR), Algorithms, SQL, NVIDIA CUDA, Regression, MySQL, Models, Artificial General Intelligence (AGI), Communication, Large Language Models (LLMs), Retrieval-augmented Generation (RAG)
NLP Expert
2022 - 2022
Prepaire Labs Limited
- Developed a drug recommender system based on a knowledge graph of interactions between drugs, genes, proteins, and diseases.
- Built a drug embedding system that can cluster drug classes without labeled data.
- Built a large-scale interaction network to predict molecule interactions between different entities.
Technologies: Python, Bioinformatics, Machine Learning, Natural Language Processing (NLP), Generative Pre-trained Transformers (GPT), Data Science, Natural Language Toolkit (NLTK), Biotechnology, Machine Learning Operations (MLOps), NVIDIA CUDA, Models, Leadership, Communication
Data Analyst | Statistician
2022 - 2022
Product Tranquility LLC
- Helped the client clean inconsistent data points in their survey data.
- Combined machine learning and data science to generate survey insights.
- Reproduced classical survey analysis techniques to analyze pricing.
Technologies: Data Science, Linear Regression, Clustering, Surveys, Survey Development & Analysis, Interoperability, Financial Modeling, Models, Communication
Data Scientist
2022 - 2022
Alaris Acquisitions, LLC
- Delivered an app to estimate the similarity between buying and selling institutions with a front-end presentation using Streamlit.
- Deployed online file synchronization for all remote users of our app using Amazon's AWS S3.
- Created a customizable interface to input and change system variables to make the app as flexible as possible while maintaining consistency and scalability.
Technologies: Algorithms, Streamlit, Python, Machine Learning, Finance, Data Mining, Data Modeling, Data Analytics, Generative Pre-trained Transformers (GPT), Natural Language Processing (NLP), Pandas, JSON, Text Mining, Algorithmic Trading, Quantitative Research, Financial Modeling, Communication
Senior Data Scientist and ML Engineer
2021 - 2022
Toptal Client
- Developed and successfully deployed a custom recommender system for one of the clients' ventures.
- Worked actively on the automatic metadata generation from users' images with transfer learning.
- Developed a churn detection model integrated with customized emails for the CS team.
- Worked actively on a semantic search algorithm based on items and user embeddings.
- Developed PowerBI dashboards to show relevant data to clients.
- Integrated our solutions with existing architectures using MLflow and email APIs.
- Developed a lead scoring algorithm based on the positive-unlabeled problem framework.
Technologies: Microsoft Power BI, Microsoft Azure, Databricks, Recommendation Systems, B2C Marketing, TensorFlow, Neural Networks, Analysis, Data Analysis, Apache Spark, Data Visualization, Statistics, Business Intelligence (BI), MLflow, Statistical Data Analysis, Analytics, Data Analytics, ETL, Atlassian, Jira, Database Analytics, Classification Algorithms, NumPy, Innovation, PySpark, Data Science, Jupyter Notebook, Data Engineering, Excel 365, Microsoft Excel, Predictive Modeling, Predictive Analytics, Data Mining, Data Modeling, Data Reporting, SQL, Python, Cloud, Deep Learning, Generative Pre-trained Transformers (GPT), Natural Language Processing (NLP), Pandas, ETL Tools, JSON, Text Mining, Technical Hiring, Code Review, Task Analysis, Interviewing, Statistical Modeling, Azure, Bash, Optical Character Recognition (OCR), Sentiment Analysis, MySQL, IT Management
Senior Data Scientist
2021 - 2021
Tecnium
- Developed the best unique product identification routine available at the time.
- Taught teammates about neural network approaches, such as autoencoders, RankNet, transformers, and embedding spaces.
- Created a product name embedding, using a transformer neural network, usable in multiple problems.
Technologies: Python 3, Microsoft Azure, Databricks, Machine Learning, Few-shot Learning, Rankings, Classification Algorithms, Azure Machine Learning, NumPy, Innovation, PySpark, Data Science, Jupyter Notebook, Data Engineering, Excel 365, Microsoft Excel, Predictive Analytics, Data Mining, Data Modeling, Data Reporting, SQL, Python, Cloud, Deep Learning, Analytics, Natural Language Processing (NLP), Generative Pre-trained Transformers (GPT), Pandas, ETL Tools, JSON, Text Mining, Statistical Modeling, Azure, Bash, Optical Character Recognition (OCR), Google Cloud ML, Regression, MySQL
Senior Data Scientist
2019 - 2021
Unimed
- Developed data-oriented models using NLP, machine learning, and transformers.
- Helped transform my current sector into a data science-oriented enterprise.
- Built useful metrics and scoring systems using embeddings and neural networks.
- Created and deployed a web crawler to generate healthcare-related disease code datasets.
- Proposed, implemented, and deployed an algorithm for future hospitalization prediction and a semantic representation for patient disease data.
- Competed as one of three finalists in the company's innovation prize awards in all the years I participated, winning the 2021 prize.
- Used time-series algorithms to predict future healthcare requirements. I used multiple neural network architectures, such as variational autoencoders, residual neural networks, and vision transformers to obtain information from medical images.
Technologies: Python 3, TensorFlow, Machine Learning, Reinforcement Learning, Generative Pre-trained Transformers (GPT), Natural Language Processing (NLP), Amazon Web Services (AWS), ARIMA, LightGBM, SVMs, Support Vector Machines (SVM), LSTM, LSTM Networks, Long Short-term Memory (LSTM), Artificial Intelligence (AI), Artificial Neural Networks (ANN), MySQL, MySQLi, SQL, Presto, Big Data, Residual Neural Networks (ResNets), Vision Transformer (ViT), Classification Algorithms, Images, Health, Azure Machine Learning, NumPy, Innovation, Transformers, Data Science, Data Engineering, Excel 365, Microsoft Excel, Predictive Modeling, Predictive Analytics, Data Mining, Data Modeling, Data Reporting, Data Analytics, Amazon SageMaker, Image Processing, Computer Vision, Python, Cloud, Deep Learning, Analytics, Pandas, ETL Tools, Text Mining, Code Review, Task Analysis, Statistical Modeling, Bash, Optical Character Recognition (OCR), Oncology & Cancer Treatment, Sentiment Analysis
Teacher
2018 - 2021
University of Medical Science
- Developed a digital teaching system, using online technologies.
- Taught students about medical research using machine learning and data science methods.
- Oversaw research projects and courses focused on recent machine learning and general technology trends.
Technologies: Python 3, Data Science, Research, Classification Algorithms, NumPy, Innovation, Jupyter Notebook, Code Review, Oncology & Cancer Treatment
Project History
Project Cindy
Cindy is a reinforcement learning agent that learns how to trade stocks based on price, volume, and other relevant data. I've built a simulator that operates on historical data to train the agent and applied it to real-time market operations. The initial version made a bridge between MetaTrader 5 and my own Python API.
In this project, I've tested many different reinforcement learning technologies, including :
• Deep-Q learning agent (DQN)
• Asynchronous actor critic agent (A3C)
• Synchronous actor critic agent (A2C)
• Proximal policy optimization (PPO)
• Convolutional Neural Networks (CNN)
In this project, I've tested many different reinforcement learning technologies, including :
• Deep-Q learning agent (DQN)
• Asynchronous actor critic agent (A3C)
• Synchronous actor critic agent (A2C)
• Proximal policy optimization (PPO)
• Convolutional Neural Networks (CNN)
Instagram Bot
I've created a fully automated Instagram automation solution as a doodle project. The robot would automatically visit profiles (web crawling), evaluate the profile against my own, estimate the probability that given it interacts with and follows the user, the user will follow it back. If the probability exceeds a certain threshold, the robot likes and comments on some posts and then follows the target user. Using this strategy, my mock account grew to 2.000 followers in around a month, after which I stopped using the bot.
Brain CT Image Alteration Detection
In this doodle project, I've set up an autoencoder to learn the reconstruction of normal brain CT images, taking a public Kaggle dataset as input, and then used the resulting embedding to obtain state-of-the-art classification results.
Automatic Labeling with Reinforcement Learning
I worked on this project for my master's degree and built a reinforcement learning agent that meta-learns based on a small labeled dataset. The agent selects chunks of data to use one of its trained classifiers and label, then adds to the labeled set and retrain other classifiers.
Electrocardiogram (ECG) Automatic Classification
https://www.kaggle.com/competitions/dcc-week-challenge-2023/overviewI won second place in this Kaggle competition for automatic electrocardiogram classification while learning from a relatively small sample of 12-lead examinations. For this project, I used a custom neural network solution.
Education
2020 - 2025
PhD in Computer Science
Federal University of Minas Gerais - Belo Horizonte, Brazil
2018 - 2020
Master's Degree in Computer Science
Federal University of Minas Gerais - Belo Horizonte
2017 - 2018
Master of Business Administration in Business Administration
Dom Cabral Foundation - Belo Horizonte
2009 - 2015
Bachelor's Degree in Medicine
Federal University of Minas Gerais - Belo Horizonte
Certifications
NOVEMBER 2022 - PRESENT
AWS Academy Graduate - AWS Academy Cloud Foundations
Amazon Web Services (AWS)
JANUARY 2014 - PRESENT
Cambridge Advanced English Certificate
Cambridge Assessment English
Skills
Libraries/APIs
TensorFlow, Scikit-learn, NumPy, Pandas, PyTorch, PySpark, LSTM, Natural Language Toolkit (NLTK), Spark ML, Amazon API, SciPy
Tools
Spyder, Atlassian, Jira, Microsoft Excel, Amazon SageMaker, Named-entity Recognition (NER), Azure Machine Learning, ARIMA, Microsoft Power BI, AWS Command Line Interface (CLI), AWS IAM, Git
Languages
Python 3, Python, SQL, C++, Bash, MQL5, C#, Snowflake, TypeScript
Paradigms
ETL, Business Intelligence (BI), Interoperability, Quantitative Research
Platforms
Anaconda, Jupyter Notebook, Amazon Web Services (AWS), Databricks, Azure, NVIDIA CUDA, MacOS, Docker, Amazon EC2, AWS Lambda
Frameworks
LightGBM, LangGraph, Selenium, Presto, Apache Spark, Streamlit, Spark
Storage
MySQL, MySQLi, JSON, NoSQL, Google Cloud, Google Cloud SQL, Azure Cloud Services, Database Management
Industry Expertise
Bioinformatics, Trading Systems
Other
Business Administration, Innovation, Machine Learning, Neural Networks, Deep Neural Networks (DNNs), Deep Reinforcement Learning, Transformers, Natural Language Processing (NLP), Reinforcement Learning, English, Data Science, Autoencoders, Medical Imaging, Algorithms, Artificial Intelligence (AI), Artificial Neural Networks (ANN), Data Analysis, Analytics, Database Analytics, Data Engineering, Predictive Modeling, Predictive Analytics, Data Mining, Data Modeling, Data Reporting, ETL Tools, Text Mining, Task Analysis, Statistical Modeling, Optical Character Recognition (OCR), Graphics Processing Unit (GPU), Oncology & Cancer Treatment, Generative Pre-trained Transformers (GPT), Sentiment Analysis, Financial Modeling, Regression, Models, Communication, Large Language Models (LLMs), Research, Microsoft Azure, Few-shot Learning, Web Crawlers, Deep Learning, Classification Algorithms, Software Engineering, Computer Vision, Vision Transformer (ViT), Recurrent Neural Networks (RNNs), Long Short-term Memory (LSTM), Web Scraping, Residual Neural Networks (ResNets), Machine Learning Operations (MLOps), Data Visualization, Statistical Data Analysis, Data Analytics, Excel 365, Image Processing, Cloud, Source Code Review, Code Review, Backtesting Trading Strategies, Artificial General Intelligence (AGI), Generative Adversarial Networks (GANs), Leadership, Retrieval-augmented Generation (RAG), IT Management, LangChain, Windows 10, BERT, Learning, Rankings, Dedupe.io, Convolutional Neural Networks (CNNs), Meta-learning, 3D Images, Images, Health, Medical Software, SVMs, Support Vector Machines (SVM), Variational Autoencoders (VAEs), Data Transformation, LSTM Networks, Big Data, Recommendation Systems, B2C Marketing, Analysis, Statistics, MLflow, Finance, Technical Hiring, Interviewing, Linear Regression, Clustering, Surveys, Survey Development & Analysis, Biotechnology, Genetic Algorithms, Mathematics, Algorithmic Trading, GPU Computing, Google Cloud ML, FastAPI, Image Generation, Cognitive Search, Classification, VM, Deduplication, Molecular Docking, Spanish, AI Agents, LLM Agents, Agentic Workflow Design
Collaboration That Works
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
1
Share your needs
Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2
Choose your talent
Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3
Start your risk-free talent trial
Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.
Top talent is in high demand.
Start hiring