Fernando Barscevicius

Building the ML and Data Science function from scratch, reporting directly to the CEO and CTO. I design the MLOps infrastructure, fine-tune language models for healthcare document understanding, and build risk prediction systems. Working with LLMs for data labeling and model monitoring. The autonomy to build combined with the pressure to ship fast in a competitive market makes this challenging and rewarding.

Led a team of 7 data scientists building propensity models for CRM and product recommendations. Created Python libraries and templates that cut model development time from 12 weeks to 4. Focused heavily on code reviews, 1:1s, and knowledge sharing to keep the team cohesive and shipping quality work.

Owned NLP systems for customer support: auto-reply and intelligent ticket routing. Prioritized data quality over model complexity and integrated clickstream data that became a key resource for other ML teams. Reduced support costs by 10% and wait times by 20%.

At a small startup, I wore many hats: ML, product decisions, sales calls, backend work. Built credit risk models from mobile transaction data, introduced Scrum to speed up development cycles, and automated backtesting pipelines to cut delivery time from 1 week to 1 day.

Built search ranking and recommendation systems to improve user experience. Created an NLP fraud detection model and an image comparison system to catch duplicate property listings. Worked closely with engineers and product teams to ship scalable ML solutions.

My first ML role. Built recommender systems for retail clients and automated analytical workflows for statisticians. Focused on understanding business problems and delivering solutions that moved metrics.

Joined a mentorship program to learn from experienced data scientists while doing pro-bono work for US startups. Analyzed health sensor data from sleep monitors to measure the impact of a light-based sleep intervention. Built models to predict flu outbreaks across the US using public health data.

Ongoing 45-hour professional development consisting of immersive problem-based learning experiences, mentorship sessions, and hands-on case resolutions, facilitated by some of Brazil's leading data executives, including CTOs and high-level executives.

A neuroscientist with interdisciplinary training can occupy a research role, or be hired by tech industries to analyze or assist in the implementation of errors and fraud detection systems, pattern identification, adaptive analysis of information in robotics, data mining, unsupervised machine learning, language analysis and synthesis, image interpretation, proactive software, among other applications. (source: http://neuro.ufabc.edu.br/teaching/undergraduate-program)

In the Bachelor of Science and Technology students have the opportunity to work and develop interdisciplinary themes, through a methodology that encourages the investigative posture, stimulates research and consequent scientific production, which provides the necessary means to trigger the process of continuous learning in the course of their future academic and professional life. (source: http://prograd.ufabc.edu.br/bct)

freeCodeCamp is a non-profit organization that consists of an interactive learning web platform, an online community forum, chat rooms, Medium publications and local organizations that intend to make learning web development accessible to anyone. Beginning with tutorials that introduce students to HTML, CSS and JavaScript, students progress to project assignments that they must complete either alone or in pairs. Upon completion of all project tasks, students are partnered with other nonprofits to build web applications, giving the students practical development experience.