Jonathan Oliveira

About

Building AI systems that hold up inside real institutions.

My work has repeatedly sat inside institutions where software has to be dependable, explainable, and useful to people under time pressure. That combination has shaped how I design AI systems: less as isolated models, more as services and workflows that have to hold up in production.

How I work

I work on the boundary between model capability and operational reality: the point where APIs, data flow, evaluation, and user trust matter as much as the model itself.

Most of my recent work has happened in legal-tech and public-sector settings, where adoption depends on reliability, explainability, and operational fit. That has shaped the way I design systems: less as isolated models, more as workflows and services that have to survive integration, governance, and real user pressure.

I care about the interface between model behavior and delivery quality. That usually means working across APIs, evaluation, data flow, retraining, and the people who ultimately need the system to help them do their work.

Focus areas

  • AI delivery for legal and public-sector workflows
  • Backend and platform boundaries for NLP, retrieval, and model-serving systems
  • MLOps foundations that make updates easier to ship and explain

Working style

  • Start with the workflow and the people using it, not with model novelty
  • Prefer explicit metrics, inspectable interfaces, and visible tradeoffs
  • Treat release safety, traceability, and feedback loops as product requirements

Experience timeline

Roles and teams where I built that delivery pattern.

The strongest signal in my profile is repeated delivery in environments where AI has to work for institutions, not just for demos.

TTY2000 | TRF1

Jan 2024 - Present

Senior Machine Learning Engineer

Brasília, Brazil

Senior machine-learning engineer leading modernization of judicial NLP services at TRF1, spanning legacy-model refactors, orchestration, APIs, retraining flows, and analyst-facing delivery in a large federal court environment.

  • Refactored legacy ML systems and reduced processing time by 25%.
  • Delivered analyst-facing Django APIs consumed by more than 500 internal users.
  • Automated ETL, retraining, and release flows with Airflow and CI/CD.
  • Python
  • Django
  • PostgreSQL
  • Airflow
  • DVC
View case study

AI.Lab UnB

Mar 2021 - May 2024

Machine Learning Engineer

Brasília, Brazil

Machine learning engineer at AI.Lab UnB delivering legal-tech R and D and production ML systems across CNJ/PNUD, PGDF, TST, and TRF1 initiatives, connecting stakeholder discovery, NLP, APIs, MLOps, and technical leadership.

  • Contributed to PEDRO, OSIRIS, SABIA, and ALEI across legal and institutional programs.
  • Built supervised, unsupervised, and semi-supervised models for production systems.
  • Standardized MLOps patterns for experiment tracking, dataset versioning, and repeatable delivery.
  • Python
  • FastAPI
  • Flask
  • MLflow
  • DVC
View case study

Arvvo Tecnologia

Jan 2019 - Mar 2020

Software Engineer Intern

Brasília, Brazil

Started in full-stack delivery, building internal products with Node.js, Vue.js, and MongoDB in a consulting environment that shaped my backend, API, and team-delivery fundamentals.

  • Built responsive internal interfaces with Vue.js and Angular.
  • Implemented REST services in Node.js and Express.
  • Worked in Git-based review workflows across frontend and backend.
  • Node.js
  • Vue.js
  • Angular
  • MongoDB
  • Express

High school project

A high-school project that still says something useful about range.

During technical high school, I was already building hardware-oriented systems and thinking about real-world interfaces, security, and reliability.

Centro de Ensino Médio Integrado do Gama (CEMI)

Feb 2016 - Oct 2017

Biometric Payment System for Public Transport

Brasília, Brazil

Built during technical high school at CEMI, this Arduino and IoT prototype explored biometric bus-fare payment as a way to replace physical tickets and cards with fingerprint-based authentication.

  • Integrated biometric authentication with a fingerprint scanner.
  • Processed fare validation on Arduino hardware with embedded C.
  • Explored IoT connectivity, RFID, and secure handling of biometric data.
  • C
  • Arduino
  • IoT
  • RFID
  • Biometric Scanner

Working principles

  • A useful AI system is one that survives integration, governance, and operator feedback.
  • Delivery quality comes from stable APIs, transparent evaluation, and repeatable operations.
  • The best architecture is usually the one a team can still reason about six months later.

Patterns I keep seeing

  • Legal and public-sector AI work rewards traceability more than novelty.
  • Adoption depends on workflow fit, not just offline model metrics.
  • Teams move faster when retraining and release steps are visible instead of implicit.

Next step

The project pages go deeper into the systems themselves.

If you want to evaluate the technical side of this background, go straight to the case studies. If you want to talk about roles or collaboration, use the contact page.

View Projects Contact