Ian Fukushima

I am currently working at Blue Orange Digital as a Machine Learning Engineer. I am building a Generative AI platform that encompass framework for evaluating LLMs, an AI Gateway to provide unified API and safe access to different models, as well as working with teams to succesufully deploy applications. I have also acted as a consultant, deploying ML and AI solutions to the gaming and automative industries.

During my time as a Data Scientist at Tabas, a Brazilian proptech, I developed a near real-time pipeline for estimating house prices for internal use, trained ML models to accurately predict property values, designed and implemented the company's data warehouse, and deployed web applications for realtors to facilitate in-depth analyses of apartments and their neighborhoods. The approach to predicting house prices was based on an ensemble of a customized KNN model and a gradient boosting model, with apartment characteristics features and geolocation features created by a decision tree, which was equivalent to clustering the area of interest in house value zones.

As a Data Scientist consultant at Indicium, I worked with a variety of clients. I developed MLOps standards for easy training and deployment of ML models, provided solutions for real estate clients by predicting demand and estimating house prices, and built a fraud detection pipeline for an e-commerce that was able to detect over $100,000 in fraud annually with minimal cost. The fraud detection algorithm was trained with Google Analytics and transaction data, which were feature engineered and stored in a feature store that fed tree-based classification models.

Master of Science Degree in Applied Economics at the University of São Paulo, a leading institution in Latin America. My research was focused on impact evaluation and social economics, with publications addressing the effects of full-time schooling and the consequences of social distancing during the COVID-19 quarantine on crime rates.