Gabriel Fritz Sluzala

- Development of ML pipelines to train and deploy a SOTA deep learning model to identify decisions, jurisprudence and legislations on brazilian legal documents; - Building APIs to integrate the ML pipelines with Atrium: A system to help lawyers on the writing of legal documents with bigger chances of winning on court; - Participating on the framing of data science objectives for projects on IoT and personalized education.

- Promoted from Data Scientist to lead 4 data scientists and 1 AI Strategist; - Development of a complex time-series ML system to generate, for Points of Sale, daily order recommendations for products based on demand prediction in order to avoid missing products on the store’s stock; - Development of ML projects on the retail sector in the order of 4mi BRL (~1mi USD), generating business answers, automating analysis and, developing ML models; - BeHy speaker: BeHy is a course directed to stakeholders (C-levels, directors, VP’s) with the purpose to technically educate them about AI and its impact in today’s world. As a speaker, I’m responsible to prepare the technical part of the course material and present them to stakeholders in an intuitive way, so they can grasp the fundamental concepts about AI; - Co-development of business plans and pitch decks for new AI Ventures (most of them related to retail); - Participation on the technical cooperation team, responsible to present and discuss business cases, and to develop new management tools to help in key stages of data science projects life cycle.

- Worked in the early stages of Kunumi’s COVID-19 death toll prediction platform. The team was composed by researchers, data scientists, ML engineers and domain specialists. My role in this project was, having the validated researchers prototypes at hand, to structure the production pipelines for feature engineering, modelling and results reporting. Besides that, I contributed to the project with suggestions on how to model the problem, so we would have the best interpretability of the results. The pipelines were structured using Kedro. The most recent version of this project can be found here: www.covid-19.kunumi.com; - Development and deployment of data and ML pipelines in environments like Databricks and Kubernetes; - Development of a Python library to automate the process of clustering arbitrary datasets, while reaching the best clustering results (in terms of interpretation), and reporting results; - Basket analysis using complex networks to answer questions like: Which products are the most important?, Which products connect different groups of clients?, What are the products that most explain some segment of clients?, Which products could be sold together in order to leverage sales?