Location: Brasília, Brazil (GMT-3)
Email: gabrielziegler3@gmail.com
GitHub: gabrielziegler3
LinkedIn: gabrielziegler
Experienced Software and Machine Learning Engineer with a proven track record of designing and deploying ML systems to production. My expertise of over 5 years includes the domains of PDF Processing, Speech Processing, NLP, and Fraud Detection.
Experiences
Senior Software Engineer ♦️ 
Datia
February 2024 - Present - Stockholm, Sweden (Remote)
Tech: Python, AWS, LLMs, Testing, OCR, PDF parsing
Python Engineer ♦️ 
Turing.com
August 2022 - February 2024 - California, USA (Remote)
Python developer for one of the biggest AI research companies.
Tech: Python, AWS, LLMs, Testing, Pytorch.
Machine Learning Engineer ♦️ 
Itaú Unibanco
March 2021 - August 2022 - São Paulo, Brazil (Remote)
ML Engineer with experience in designing, implementing, deploying, and monitoring speech ML systems. Worked with Data Science and MLOps teams to deliver data-intensive systems.
- Improved speaker verification model by ~20% and speaker diarization model by ~30% on noisy datasets
- Deployed APIs and end-to-end batch-processing Speech Recognition System
- Experienced in using AWS infrastructure, including SQS, SNS, S3, Sagemaker, ECS Fargate, ECR, EFS, Lambda, and API Gateway
Tech: API Gateway, IAM, KMS, ECS, EFS, CodeBuild, CodePipeline, CloudWatch, Fargate, Lambda, Sagemaker, SQS, SNS, S3, Python, Pytorch, Hugging Face Transformers, Docker, Flask, Linux.
Data Scientist ♦️ 
Banco do Brasil
March 2020 - September 2020 - Brasília (Remote)
Responsible for ETL pipelines, EDA, Insights Reports and client-focused Machine Learning Models for Classification, Regression, and Clustering.
- Improved the credit limit model by 10 percentage points in F1-Score.
Tech: Python, SQL, H2O AutoML, SAS, IBM DB2.
Data Scientist ♦️ 
Vert
April 2019 - July 2019 - Brasília, Brazil
Responsible for creating a dashboard in SAS Viya with data from national transport traffic.
Tech: SAS Viya, SQL, Python.
Facebook F8 Hackathon Finalists 2019
April 2019 - April 2019 - San Jose, California
Selected as Top 8 Finalist out of 55 teams in the annual Facebook F8 Hackathon. Our team pitched to Mark Zuckerberg personally being primarily responsible for using Deep Learning models for text summarization and translation.
Machine Learning Researcher & Developer (NLP) ♦️ 
AI Lab
March 2018 - April 2019 - Brasilia, Brazil
Responsible for EDA, ETL and development of the multiclass and multilabel (over 1000s of classes) ML model for document classification for the supreme court in the biggest AI project in the Brazilian Judiciary.
- Model performance F1 score > 80%, performing way above the human performance evaluated for the same task.
Tech: Python, Linux, RabbitMQ, PostgreSQL, XGBoost, SpaCy, NLTK, Docker, Git.
Education
MSc Data Science ♦️ 
Aalto University
September 2022 - July 2024 - Helsinki, Finland
Scholar in the EIT Digital double master's degree in Data Science & Artificial Intelligence.
MSc Data Science ♦️ 
Eindhoven University of Technology
September 2022 - July 2024 - Eindhoven, Netherlands
Scholar in the EIT Digital double master's degree in Data Science & Artificial Intelligence.
Summer School ♦️ 
Technische Universitët München
July 2023 - August 2023 - Munich, Germany
Collaborated with Siemens on creating 'EmissionAware': a new software business to ease access to carbon-emission within the supply chain for forecasting and reducing Scope 3 emissions.
Bachelor of Software Engineering (4.13/5.0) ♦️ 
University of Brasília
August 2015 - May 2021 - Brasília, Brazil
Thesis: Context-Dependent Probabilistic Prior Information for MRI Reconstruction
Exchange Student ♦️ 
Rangitoto College
January 2014 - December 2014 - Auckland, New Zealand
Publications & Awards
- Dropping Incomplete Records is (not so) Straightforward - IDA 2023: Advances in Intelligent Data Analysis XXI
- Document classification using a Bi-LSTM to unclog Brazil’s Supreme Court - NeurIPS 2018 Workshop on Machine Learning for Developing World (ML4D)
- Document type classification for Brazil’s Supreme Court using a Convolutional Neural Network - ICOFCS 2018 Selected as Best Paper.
Languages
- Portuguese: Native
- English: Fluent - C1 TOEFL IBT 102
- Spanish: Limited working proficiency