RafaelloAI/ML Engineer - Computer Vision Engineer - Data Analyst - Data Scientist
Portfolio
About
4 years of experience in building AI/ML models using scikit-learn, PyTorch/TensorFlow,
performing data manipulation with pandas and NumPy in Python and R, and deploying AI/ML
models with Flask and FAST API.
Monte Carlo Reinforcement Learning for Simple Games - Have you ever thought about training your own recommendation system or building your own robot or creating your own chess AI that can beat even the most experienced player? Reinforcement Learning is what you need. In this project, you will explore the basics of Reinforcement Learning and Monte Carlo Method. You will learn about training your own agent to navigate and succeed in simple and complex games/environments. Discover better ways to train your agent and how to work with the environment. Link - https://cognitiveclass.ai/courses/course-v1:IBM+GPXX0MNFRU+v1
Client:
IBM Skills Network
Date:
March 2023
Service:
Artificial Intelligence
App:
Project 5
This project analyzes greenhouse gas (GHG) emissions from industrial chemical processes in the United States using data from the EPA’s Greenhouse Gas Reporting Program (GHGRP). It focuses specifically on emissions from chemicals used or produced in Production & Treatment (P&T) operations at large facilities. The analysis identifies high impact greenhouse gases such as hydrofluorocarbons (HFCs), perfluorocarbons (PFCs) and other fluorinated gases and evaluates their contribution to total emissions.
To standardize the environmental impact across different chemicals, the project integrates the Climatiq API to convert reported chemical quantities into their CO₂-equivalent (CO₂e) values. This enables accurate comparison of chemical-level emissions and highlights the most significant contributors to overall GHG output.
In addition, the project includes a Streamlit web application that allows users to upload their own datasets, automatically calculate CO₂e emissions through the Climatiq API, and visualize emission profiles across chemicals or facilities
Know Your Worth, Build A Job Salary Estimator - Want to know how much a data science job pays? Use our cutting-edge model that utilizes text mining and BERT to estimate salary from job posting descriptions. Simply copy and paste the job description to get your estimate. Say goodbye to hours of research; our model will give you the information you need quickly and easily. Link - https://cognitiveclass.ai/courses/course-v1:IBMSkillsNetwork+GPXX086PEN+v1
This project orchestrate multi-agent AI workflows such as (invoice anomalies, billing discrepancies, professional dispute resolution) using CrewAI framework and integrates LLaMA3, GP-4, Claude models for advance decision-making and streamline finance operations with less human intervention
This project focus on building secure system that classify fradulent transactions using SVM, Decision Tree models using sklearn library in Python and also several visualizations(Pie chart, Histogram) are demonstrated using matplotlib library to idenify transaction amount and class ratio. Decision tree performance is build and evaluted with ROC-AUC score with 93% by ranking top positive class over negative class and SVM's performance is evaluated with ROC-AUC score and hinge loss: 0.5. Overall, both models performs well to rank fradulent transactions over non-fradulent which helps analyts to act promptly for fradulent transactions. Models are deployed using Streamlit in Python to provide interactive web interface
Built an interactive Tableau dashboard using 84K+ customer records to analyze loyalty trends, sales performance, and coupon response. Created map, bubble, stacked bar, scatter, and combo charts to visualize revenue by region, loyalty status, and product line. Identified that Smart Electronics and TV & Video Gaming generated the highest customer lifetime value and coupon campaigns boosted sales among Bronze members