Name: Jonathan Oliveira
Role: Senior AI Engineer | Software Engineer
Positioning: Builds production-grade LLM systems, APIs, and ML platforms
Summary: AI and software engineer focused on NLP, LLM systems, and production ML platforms. Over five years of experience shipping models and APIs into large public-sector environments, improving case classification, retrieval, retraining, and governance workflows with measurable operational impact.
Location: Brasilia, Brazil
Work Model: Remote; open to relocation depending on conditions.
Compensation:
- Minimum hourly rate: USD 30/hour

Roles Fit:
- Senior AI Engineer
- Senior Software Engineer for AI Platforms
- Applied AI Engineer
- LLM Platform Engineer

Industries:
- Legal Tech
- Public Sector AI
- Enterprise AI

Core Skills:
- Programming and Data: Python, Node.js, SQL, Pandas, scikit-learn, TensorFlow, PyTorch, SQLAlchemy
- Backend: FastAPI, Flask, Django, REST API design, Nginx, Microservices
- MLOps and Infrastructure: MLflow, DVC, BentoML, Airflow, Docker, GitLab CI/CD, Kubernetes, Terraform
- LLMs and Retrieval: RAG, Multi-agent systems, Prompt engineering, Fine-tuning, Knowledge graphs, Context caching, LangChain, LangGraph, LlamaIndex, Ollama, Qdrant, Hugging Face Transformers
- Datastores and Cloud: PostgreSQL, MySQL, Oracle, SQL Server, SQLite, MongoDB, AWS, AWS, GCP
- Document intelligence systems for structured extraction from unstructured PDFs and legal documents
- Production RAG architecture with retrieval tuning, chunking strategy, and grounded response design
- Conversational AI design for decision-support workflows with traceable context and response control
- LLM evaluation pipelines using ground-truth and LLM-as-a-judge methodologies
- Agent orchestration patterns for long-running AI workflows using queue workers and API services
- Model-provider abstraction layers for consistent multi-model integration in production

Key Outcomes:
- 85% Less manual classification work at TRF1
- Weeks to days Retraining cycles shortened
- 30+ Precedent categories surfaced
- National-scale Programs across courts and agencies

Additional Delivery Achievements:
- Improved structured extraction accuracy by up to 60% through prompt optimization and evaluation-driven iteration
- Increased release confidence by implementing factuality and hallucination quality gates before production rollout
- Scaled long-running AI execution with worker-based orchestration and API coordination in containerized environments
- Reduced integration friction by standardizing model access and service boundaries across AI workflow components

Experience Highlights:
- TTY2000 | TRF1 | Senior Machine Learning Engineer | Jan 2024 - Present | 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.
- AI.Lab UnB | Machine Learning Engineer | Mar 2021 - May 2024 | 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.
- Arvvo Tecnologia | Software Engineer Intern | Jan 2019 - Mar 2020 | 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.

Initiatives:
- Projeto Osíris - PGDF | Data Scientist | Associated with AI.Lab UnB | May 2023 - May 2024 | Research and delivery initiative for PGDF focused on AI support for fiscal execution procedures, combining model automation, production APIs, active learning, and early LLM exploration for legal-fiscal operations.
- Projeto SABIÁ - TST | AI Research Team Lead | May 2023 - May 2024 | Strategic R and D initiative for the Superior Labor Court focused on evolving the Bem-Te-Vi platform through jurisprudence analysis, process clustering, and Long Life Machine Learning patterns.
- Projeto PEDRO - CNJ / PNUD | Data Scientist | Associated with AI.Lab UnB | Jul 2022 - May 2023 | National-scale AI initiative for qualified precedent discovery across STJ and STF, designed to systematize jurisprudential patterns and expose them through institution-ready analytical services.
- Projeto ALEI - TRF1 | Data Scientist & Backend Developer | Associated with AI.Lab UnB | Mar 2021 - Jul 2022 | AI-driven initiative for second-instance chambers at TRF1, combining backend architecture, machine learning, document analysis, and direct enablement of legal staff.

Initiative Achievements:
- Projeto Osíris - PGDF: Designed an active-learning loop that enabled continuous model improvement with minimal manual labeling.
- Projeto Osíris - PGDF: Successfully deployed machine-learning models into production APIs consumed by internal PGDF systems.
- Projeto SABIÁ - TST: Delivered functional unsupervised NLP pipelines that enabled clustering and exploration of legal case data.
- Projeto SABIÁ - TST: Initiated and validated a prototype LLML workflow for future AI adaptability through user feedback.
- Projeto PEDRO - CNJ / PNUD: Designed a topic-modeling pipeline that enabled identification of more than 30 precedent categories based on semantic similarity.
- Projeto PEDRO - CNJ / PNUD: Led integration of AI components with judicial data systems, expanding CNJ analytical capabilities.
- Projeto ALEI - TRF1: Successfully led a hybrid AI-backend team while balancing mentorship and delivery.
- Projeto ALEI - TRF1: Built a semi-automated data-analysis pipeline that reduced manual work.

Selected Projects:
- PEDRO Precedent Discovery Platform | https://jonathanoliveira.dev/projects/pedro-precedent-discovery-platform | National-scale precedent discovery initiative for CNJ and PNUD, combining FastAPI services, unsupervised NLP, semantic grouping, and governed experimentation to systematize qualified precedents from Brazil's highest courts.
- OSIRIS Legal-Fiscal AI Workflows | https://jonathanoliveira.dev/projects/osiris-legal-fiscal-ai-workflows | AI delivery for PGDF legal-fiscal operations, spanning production APIs, supervised and semi-supervised models, active learning, and early LLM exploration for document-heavy institutional workflows.
- SABIA Bem-Te-Vi LLML Research | https://jonathanoliveira.dev/projects/sabia-bem-te-vi-llml-research | Strategic R and D work for the Superior Labor Court focused on evolving the Bem-Te-Vi platform with unsupervised NLP, clustering workflows, and Long Life Machine Learning patterns designed for continuous adaptation.
- ALEI TRF1 Legal Analysis Platform | https://jonathanoliveira.dev/projects/alei-trf1-legal-analysis-platform | AI-driven initiative for TRF1 that brought together backend delivery, unsupervised machine learning, ETL, and court-staff enablement to improve legal analysis and case-management workflows.

High School Projects:
- Biometric Payment System for Public Transport | Feb 2016 - Oct 2017 | 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.

Technical Writing:
- Production RAG Systems Need More Than Retrieval Demos | https://jonathanoliveira.dev/blog/production-rag-systems | A production RAG system should be treated as a retrieval and evaluation pipeline with explicit failure modes, not as a prompt wrapper around a vector store.
- LLM Evaluation in Production Starts With Explicit Failure Modes | https://jonathanoliveira.dev/blog/llm-evaluation-in-production | Evaluation is most useful when it reflects the failures a system can actually produce in production: missing context, wrong retrieval, incorrect tool use, unstable outputs, and unhelpful responses.
- Scaling ML Pipelines Means Reducing Hidden Manual Work | https://jonathanoliveira.dev/blog/scaling-ml-pipelines | ML pipelines usually fail to scale because they depend on undocumented manual steps around data preparation, retraining, packaging, and release coordination.

Contact:
- Email: hello@jonathanoliveira.dev | mailto:hello@jonathanoliveira.dev
- LinkedIn: linkedin.com/in/jonathan-jorge-oliveira | https://www.linkedin.com/in/jonathan-jorge-oliveira
- Website: jonathanoliveira.dev | https://jonathanoliveira.dev
- GitHub: github.com/Jonathan-Oliveira | https://github.com/Jonathan-Oliveira

Key URLs:
- Home: https://jonathanoliveira.dev/
- Projects: https://jonathanoliveira.dev/projects
- Blog: https://jonathanoliveira.dev/blog
- Resume: https://jonathanoliveira.dev/cv
- Recruiter PDF (EN): https://jonathanoliveira.dev/downloads/jonathan-jorge-oliveira-cv.pdf
- Recruiter PDF (PT-BR): https://jonathanoliveira.dev/downloads/jonathan-jorge-oliveira-cv-pt-br.pdf
- LLM TXT: https://jonathanoliveira.dev/llm.txt
- Contact: https://jonathanoliveira.dev/contact