Joao Diogo de Oliveira
Verified Expert in Engineering
Machine Learning Engineer and Developer
Joao is an AI/ML engineer with more than 14 years of experience at Fortune 100 companies like Procter & Gamble and Hearst and startups in the healthcare, energy, and finance industries. Joao holds a master's degree in computer engineering from the University of Porto and has multiple certifications in ML and deep learning.
Python 3, PyTorch, TensorFlow, R, Machine Learning, Google Cloud Platform (GCP), Amazon Web Services (AWS)
The most amazing...
...project I've led is predicting over 300 wind and solar farms in a record time of 1.5 months.
Hearst - Technology
- Successfully developed an MVP, demonstrating the ease of replacing a legacy system (built over the course of 10+ years with an estimated cost of $1+ million) within 3 – 4 weeks.
- Steered non-technical discussions to extract the needed information to build the MVP and alignments on deliverables.
- Used generative AI (GPT 3.5, GPT 4) and other frameworks and libraries (LangChain, LlamaIndex) to extract structured data from medical labels (unstructured data), achieving up to 98% success rate.
Peyton & Greyson Solutions Inc,
- Developed an AI application for writing automatic proposals, saving at least 20% of the time from a specialized employee.
- Designed and architected the entire IT solution: a) database choice and detail; b) AWS Serverless Services; b) chose and set up the web app back-end implementation; c) API configuration; d) to complete AI model development and deployment.
- Tracked team members' development and ensured that milestones were met, from demos to critical development deliverables.
- Tailored the GPT-3 model to a specific business case successfully.
IT Engineer | Artificial Intelligence Engineer
- Developed an artificial intelligence AI project for energy prediction of solar and wind farms, summing up 2.6 GW of installed power.
- Built a model for computer vision that did face recognition.
- Created a model using computer vision to ease pneumonia detection through X-ray.
- Provided consulting services to deliver wind certification for two offshore projects with a combined predicted installed power of 2GW.
- Maintained over 20 distributed Linux servers, updating, securing, and creating key performance indicators KPIs.
Product Owner | Country Manager
- Developed AI models, including deep learning, weather forecast, and energy prediction for multiple markets.
- Performed business and data analytics for customers.
- Led the successful establishment of a European institute in Brazil.
- Managed a portfolio of clients with a combined production of over 3 GW of energy.
- Managed daily operations of the group's investments, including a shopping mall, business towers, and representation before official government bodies.
- Reversed a seven-year loss into profit by applying substantial and stable changes.
- Supervised the financial control of the construction of three towers, 12 floors each, with a total cost of R$ 43 million.
Stop the Traffik
- Analized the most underlying tech issues in a volunteer organization and proposed a plan to tackle them through a team of 11 volunteers scattered over nine countries.
- Led a team of ML/AI specialists to develop an AI model for sentiment analysis to automize the analysis of trafficking articles and classify them, removing the manual labor currently applied.
- Guided and led a team of ML/AI specialists to improve the legacy model that classified articles into relevant and non-relevant ones for the organization.
- Steered through meetings the project success and engagement to deliver the proposed outcomes to the organization. Participated in all parts of development (AI, DevOps, Python) to make sure that commitments were met and delivered.
Mercatus Center at George Mason University - Main
- Developed a long text classification for documents within 96 labels. The purpose was to use different NLP techniques to get probabilities of the three digits NAICS codes.
- Explored literature on the most advanced techniques of text classification and long text and applied them; Combined the different techniques to achieve a better result, achieving an improvement of 15% on the F1 score.
- Used AWS SageMaker to provide an effective and insightful training and inference pipeline.
- Achieved F1 scores on some categories up to 0.95-0.98 on others using different techniques increased from 0.4 to 0.7.
Procter & Gamble
- Implemented multiple line update projects across plants in France, Italy, and Spain.
- Developed cost-saving solutions and deployed them across multiple factories.
- Led technical discussions with suppliers to make sure they would meet the requirements.
Supply Chain Leader
Procter & Gamble
- Led the design and implementation of a global pilot project to remodel the company's logistics sector.
- Found a solution to complex problems of inventory costs, achieving a reduction from $12 million to $7 million.
- Participated in creating an internal cross-docking supply chain prototype, resulting in yearly savings of $2 million.
- Coached, guided, and coordinated the work of multiple team members.
NLP in Healthcare | Score Clinical Patient Noteshttps://www.kaggle.com/c/nbme-score-clinical-patient-notes
CV: X-ray Pneumonia Detectionhttps://github.com/joao-d-oliveira/X-Ray_PneumoniaDetection
Power Generation Forecast for Wind and Solar Farms
Computer Vision - Face detection
Developing AI Automated Proposal Generation
CV: Image Captioning - Identifying Objects and Writing Caption
Email NLP/NLU/NER Analysis
Surgery Assistance Software
I designed and implemented the architecture of the software, achieving an MVP.
Python 3, SQL, Python, R, Python 2, C++
Scikit-learn, Pandas, LSTM, Matplotlib, Keras, PyTorch, TensorFlow, OpenCV, PyTorch Lightning, FFmpeg
GitHub, Amazon SageMaker, You Only Look Once (YOLO), NLPP
Data Science, Agile, DevOps, Anomaly Detection
Linux, Amazon Web Services (AWS), Jupyter Notebook, Kubernetes, Docker, Google Cloud Platform (GCP), Backendless, AWS Lambda
Data Pipelines, PostgreSQL, MySQL
Machine Learning, Deep Learning, Data Structures, Artificial Intelligence (AI), Algorithms, Team Leadership, Project Design & Management, Computer Vision, BERT, Natural Language Processing (NLP), Deep Neural Networks, Datasets, Language Models, Image Processing, Hardware, Models, AI Programming, Data Processing Automation, Real Estate, ARIMA, ARIMA Models, Supply Chain Management, Supply Chain Optimization, Forecasting, Information Extraction, Energy, Neural Networks, Regression Modeling, Data Processing, Data Transformation, CSV, Data Analysis, GPT, Generative Pre-trained Transformers (GPT), Back-end, Generative Pre-trained Transformer 3 (GPT-3), OpenAI GPT-4 API, Workshop Facilitation, Analytics, Convolutional Neural Networks, Sentiment Analysis, Data Analytics, Process Management, Logistics, Statistics, Computer Vision Algorithms, Data Visualization, Big Data Architecture, Machine Learning Operations (MLOps), Generative Adversarial Networks (GANs), OpenAI, DaVinci, Large Language Models (LLMs), SARIMA, Natural Language Understanding (NLU), Hugging Face, Cloud Platforms, Early-stage Startups, Generative Artificial Intelligence (GenAI), Web Development, Word Embedding, ChatGPT, OpenAI GPT-3 API, API Integration, Speech Recognition, Scraping, Facial Recognition, Image Recognition, Finance, Quantum Computing, Healthcare IT, Deep Reinforcement Learning, APIs, Object Detection, Generative Models, AI Design, Amazon RDS, Image Generation, CTO, Transformers, IBM Cloud, Prompt Engineering
Master's Degree in Computer Science
University of Porto - Porto, Portugal
Exchange Program Coursework Toward Master's Degree in Computer Science
Delft University of Technology - Delft, Netherlands
Quantum Excellence Certificate
IBM | Qiskit Global Summer School 2022
AI for Healthcare
Deep Reinforcement Learning
Advanced Computer Vision - Machine Learning