Open to opportunities

Beatriz Maia

@beatrizmaia

Machine Learning Engineer specializing in LLM orchestration and agentic AI.

I am a Machine Learning Engineer focused on LLM orchestration and agentic AI, building production-grade systems that integrate models from OpenAI, Google, and Anthropic into scalable backend infrastructures.

I design agentic workflows, fine-tune large language models, and implement RAG and CAG systems; I have engineered SDK features, developed APIs for transcription and RAG-as-a-service, and automated evaluation pipelines using LLM-as-a-judge.

My background includes deep learning research published in IEEE, hands-on MLOps experience with multi-GPU training (DeepSpeed, FSDP, Accelerate), and deployment work across GCP, Azure, and AWS, improving operational efficiency and reducing manual processes.

I bring bilingual communication skills, multicultural adaptability, and a strong combination of academic research and practical engineering to help teams productionize advanced generative AI solutions.

Experience

Work history, roles, and key accomplishments

Current

Machine Learning Engineer

Current

Cadastra

Apr 2025 - Present (10 months)

Maintains a GenAI Gateway integrating OpenAI, Google, and Anthropic via Azure/VertexAI and engineered SDK features for model-switching, reducing a manual 1-hour data update to 5 minutes through an autonomous GCP pipeline.

fastAPI GCP PostgreSQL SQLAlchemy LiteLLM RAG Unit Testing

Machine Learning Engineer

Docket

Oct 2024 - Nov 2024 (1 month)

Fine-tuned multimodal models for NER on legal deeds and used Google Cloud Vision and AWS S3 to preprocess high-variance notary documents for property ownership data extraction.

NER AWS S3 data preprocessing PyTorch

Machine Learning Engineer

Weni

Nov 2023 - Oct 2024 (11 months)

Fine-tuned and aligned LLMs up to 70B parameters using DeepSpeed, FSDP and RLHF/DPO, applied AWQ quantization for production efficiency and advanced the team's MLOps via multi-GPU semi-autonomous pipelines.

PyTorch DeepSpeed Fsdp Hugging Face RLHF Pipeline DPO MLOps RunPod

Researcher

Bio-inspired Computer and Engineering Lab

Aug 2020 - Oct 2024 (4 years 2 months)

Led research on Transformers and Few-Shot Learning for digital pathology, publishing in IEEE Expert Systems with Applications (2023), and investigated VAEs for medical feature extraction.

Transformers Few Shot Learning VAES Medical Imaging Research Scientific Writing Python Deep Learning

Machine Learning Engineer

Eldorado Research Institute

Jun 2022 - Oct 2023 (1 year 4 months)

Developed hybrid OCR pipelines with OpenCV/Tesseract and YOLO, built NER pipelines with Hugging Face, and curated secure Python Docker images while contributing to MLOps and XAI research groups.

OpenCV Tesseract YOLO Hugging Face NER Docker MLOps Explainable AI Python

Data Scientist

VERT Transformacao em TI

Jun 2021 - Jun 2022 (1 year)

Led client engagements and risk analysis for government bidding and DETRAN-ES projects, architected ETL processes in SAS and implemented web scrapers for automated public data collection.

SAS ETL PostgreSQL SQLAlchemy BeautifulSoup Selenium Risk Analysis Client Management

Education

Degrees, certifications, and relevant coursework

Federal University of Espírito Santo

Bachelor's Degree, Computer Science

2019 -

Activities and societies: Coursework emphasis on Machine Learning and Software Engineering; involvement in research lab (LABCIN) and voluntary teaching activities.

Bachelor's Degree in Computer Science with emphasis on Machine Learning and Software Engineering; coursework completed in 2024, degree expected March 2026.