Caylent

Senior Machine Learning Engineer

SKF Group

Building AI for machine condition monitoring, vibration analysis, and predictive maintenance.

Machine Learning Engineer

Shake

•  Built classification models with NLP, small LLMs (GenAI) to compare website similarity for web data aggregation for brand management, participating in a 30% increase in revenue. 
• Fine-tuned LLMs like miniLM and BERT using HuggingFace, PyTorch, and PEFT (LoRA, QLoRA) for web data aggregation, resulting in a 5% accuracy increase.
• Used Vector Databases (ElasticSearch) with embedding models to store and retrieve relevant webpages for small LLMs, resulting in a 10% speed improvement.
• Built their first MLOps pipeline using AWS S3 and AWS SageMaker to use GPUs in our architecture, resulting in 3x times faster inference time.

Data Scientist

BTG Pactual

• Increase in 50% in the development speed for the app Chatbot after I’ve built a Topic Modelling tool that tagged customers topics.
• Estimated increase of 10% in customer call acceptance after I’ve developed a
Recommendation Systems model to predict the best time to call.
• Increase in 100% of directed market campaign for airports marketing using
geospatial analysis.

Oncase

• Worked on the creation of a complete data product that ingested at least 30% faster many different databases, performing full ETL procedure in a parallelized manner using AWS.
• Created data pipelines in order to enrich the data and extract insights for the clients in the context of market research of their products.

Data Engineer

NavalPort

• Pulling data from the database, I modeled their data using Queueing Theory in order to evaluate statistical relevant information and forecast precise navigational and shipping processes.
• Created from scratch data pipelines using Postgres and GIS to extract and explore navigational coordinates in harbors.
• Developed a dashboard to interactively visualize and explore the different outputs from the previous steps.

Universidade Federal de Pernambuco

• I led the construction of the dashboard to carry out interactive reports in addition to visualisation and inspection of fund data, which improved by 50% the decision flow of RECIPREV analysts.
• Through the use of multivariate analysis techniques such as PCA, I proposed new valuable metrics for RECIPREV’s manager, which improved decision making.
• I've also implemented advanced statistical techniques and metrics to analyse and model financial data (GARCH, Markowitz Analysis, etc.), helping to increase profitability and reduce risks in RECIPREV's investments.

• I've researched and reviewed the scientific literature in search of top current statistical models, generating a concrete document that summarises the statistical models in this area. Also, testing in never-before seen scenarios.
• I also developed the beginning of an R package as a collection of codes to advance the research in my specific topic. The package is called distkrige.

Undergraduate Researcher

Passionated about Natural Language Processing (NLP), Large Language Models (LLM) and Computer Vision, I seek to understand and disseminate digital intelligence in our societies to answer question and solve problems.

Besides my 4 years of professional experience in Data, I have a BSc in Statistics and I'm currently finishing my Master's Degree in Computer Science with a focus on Artificial Intelligence. I have also a certificate in the Computer Vision Nano-degree program by Udacity.

I'm also happily married, love coffee, video-games and training Brazilian Jiu-Jitsu (BJJ).

Tools and Knowledge:
Machine Learning
Deep Learning
Statistics
MLOps
AWS
Python, R, Julia
PyTorch
LLM
REST API
Data Engineering
Docker
Airflow
Vector Databases

Taplio

Adriel Martins's Linkedin Analytics

Adriel Martins

Want to drive more opportunities from LinkedIn?