Marcos Paulo Quintao Fernandes, Developer in Belo Horizonte - State of Minas Gerais, Brazil
Marcos is available for hire
Hire Marcos

Marcos Paulo Quintao Fernandes

Verified Expert  in Engineering

Artificial Intelligence (AI) Engineer and Developer

Belo Horizonte - State of Minas Gerais, Brazil

Toptal member since May 31, 2024

Expertise
Data MiningNLPMachine LearningArtificial IntelligenceTensorFlowPythonPyTorchElasticsearchDeep LearningLLMC++Cloud EngineeringC#
Bio

Marcos is an experienced engineer with over five years of experience tackling complex search-related challenges that blend software engineering and machine learning expertise. He excels in leveraging ranking features, developing ranking models, building recommendation systems, and extracting topics. Proficient in enhancing language models through fine-tuning, domain adaptation, model quantization, and distillation, Marcos has made significant contributions to his field.

Portfolio

Jusbrasil
Python, Elasticsearch, Topic Modeling, Natural Language Processing (NLP)...
Microsoft
C#, Machine Learning, Topic Modeling, Artificial Intelligence (AI), PyTorch...
Google
C++, Data Mining, Machine Learning, Python, Artificial Intelligence (AI)...

Experience

  • Information Retrieval - 7 years
  • TensorFlow - 5 years
  • Natural Language Processing (NLP) - 5 years
  • Python - 5 years
  • Machine Learning - 5 years
  • Software Engineering - 5 years
  • Elasticsearch - 3 years
  • Google Cloud Platform (GCP) - 3 years

Availability

Part-time

Preferred Environment

Google Cloud Platform (GCP), TensorFlow, Python, C++, Scikit-learn, Natural Language Processing (NLP)

The most amazing...

...initiatives I've headed at Jusbrasil have boosted engagement metrics by over 10%.

Work Experience

Staff Software Engineer

2022 - PRESENT
Jusbrasil
  • Headed the search-ranking team to analyze and enhance the system's ranking, driving initiatives from an A/B testing platform to learning-to-rank projects.
  • Collaborated with the product team to understand and consolidate key engagement metrics and user journeys. Used those existing metrics to build a new A/B testing system.
  • Developed a learn-to-rank pipeline that gathers user feedback to optimize the combination of ranking features, ensuring the highest quality ranking outcomes.
  • Led initiatives to deploy generative AI to create snippets for popular queries and question-answering formats.
  • Trained models to identify question-answer user queries using transformers and deployed them to handle 70 requests per second in real time, with costs below $400 per month.
  • Oversaw and trained a domain-adapted bidirectional encoder representations from transformers (BERT) model for the legal field to execute dense vector searches, integrating the retrieval-augmented generation (Legal-RAG) pipeline company-wide.
Technologies: Python, Elasticsearch, Topic Modeling, Natural Language Processing (NLP), Machine Learning, Information Retrieval, Artificial Intelligence (AI), Azure AI Studio, Generative Pre-trained Transformers (GPT), Large Language Models (LLMs), Large Language Model Operations (LLMOps), PyTorch, Data Science, GPU Computing, Open Neural Network Exchange (ONNX)

Software Engineer

2020 - 2022
Microsoft
  • Designed and executed A/B tests for various topic extraction methods, using click models and user feedback to determine the superior algorithm.
  • Designed a cost-effective architectural update and led a team of two members to enhance processing efficiency.
  • Launched this new architecture, yielding over a 20% reduction in central processing unit (CPU) costs alongside a minimal 5% increase in memory expenses.
Technologies: C#, Machine Learning, Topic Modeling, Artificial Intelligence (AI), PyTorch, Data Science

Software Engineer | Intern

2017 - 2018
Google
  • Constructed a token classification pipeline for Health Search to extract health-related entities.
  • Developed a "Is Health Related" document classification pipeline using token-level analysis, achieving over 0.95 F1 score precision across over 10 billion documents.
  • Added the is_health_related_feature to the Google Health search trigger that directs health-related searches to general search.
Technologies: C++, Data Mining, Machine Learning, Python, Artificial Intelligence (AI), Data Science

Experience

Ranking and Recommendation Model

Oversaw and deployed all initiatives to enhance the ranking of various legal documents at Jusbrasil, a legal technology firm. The company’s repository comprises over 300 million records and a dataset exceeding two terabytes.

DELIVERABLES
• Development and application of generative AI to produce key document excerpts.
• Implementation of ranking algorithms that incorporate user feedback to refine document relevance.
• Establishment of click models to track and analyze user engagement.

Education

2013 - 2018

Bachelor's Degree in Electrical Engineering

Federal University of Minas Gerais - Belo Horizonte, Brazil

Skills

Libraries/APIs

TensorFlow, PyTorch, Scikit-learn

Tools

Open Neural Network Exchange (ONNX)

Languages

Python, C++, C#

Platforms

Azure AI Studio, Google Cloud Platform (GCP)

Storage

Elasticsearch

Other

Natural Language Processing (NLP), Data Mining, Machine Learning, Information Retrieval, Artificial Intelligence (AI), Data Science, GPU Computing, Deep Learning, Topic Modeling, Generative Pre-trained Transformers (GPT), Large Language Models (LLMs), Large Language Model Operations (LLMOps), Software Engineering

Collaboration That Works

How to Work with Toptal

Toptal matches you directly with global industry experts from our network in hours—not weeks or months.

1

Share your needs

Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2

Choose your talent

Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3

Start your risk-free talent trial

Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.

Top talent is in high demand.

Start hiring