NLP Researcher

Full-time

As part of the HAILab team, I spearheaded the training of large language models (LLMs) on biomedical and clinical data. Moreover, I have been involved in all stages of data integrity, including meticulous cleansing and anonymization of sensitive patient information.

Utilizing the Vertex-AI platform, I have conducted training sessions employing techniques such as Lora, qLora, and Full, aimed at developing a model capable of generating clinical texts. My expertise encompasses proficiency in training models, using a spectrum from single A100 GPU utilization to harnessing parallelism across multiple GPUs. Additionally, I have actively participated in the complete model inference pipeline, ensuring a smooth transition from training to deployment.

Technical cases:

Pre-training LLama2 7B:

Using the LLama-Factory library in conjunction with Huggingface Accelerate, I conducted pre-training of the LLama 7B base model with new data(Clinical and Biomedical)...

Fine-tuning:

After injecting new knowledge into my model, I performed fine-tuning on biomedical QA datasets using Supervised Fine-Tuning techniques...

Inference:

I actively participated in the model inference process... Combining various techniques and evaluating the model using benchmark approaches such as the mmlu dataset and creating scripts to test multiple models simultaneously and assess response performance.

SEO Data Scientist

Internship

As a member of the Data team specializing in SEO, I have actively participated in projects centered around API integration, web crawler-based data collection, and report automation. Additionally, I developed tools using Large Language Models (LLMs). As a matter of fact, I executed automation using Selenium in EC2 instances, aiming to extract data from various sources such as Google Search Console, Google Trends, site: searches, robots.txt files, and logs. The data obtained was then securely stored on AWS S3.

Technical cases:

GPT API:

I developed an automated dashboard that utilizes the GPT-3 model to generate text based on data gathered from Crawlers and applications such as ScreamingFrog, Ahrefs, and GA4. I designed and implemented the entire data pipeline and API trigger. Currently, this dashboard is being utilized by 32 clients.

Automation using AWS:

I implemented automation using Selenium on EC2 instances to scrape data from various sources such as Google Search Console, Google Trends, site: searches, robots.txt files, and logs. The scraped data was then securely stored in AWS S3

RAG Q/A:

Leveraging cutting-edge LLM tools, I triumphed at the company hackathon. My project used Langchain, Chromadb, Streamlit, and ChatGPT to create an application where it was possible to ask questions/answer to the company's training PDFs. I constructed a vector database with diverse PDFs on departments, roles, timekeeping, and other new hire resources. This project, showcasing my expertise in LLM and vector database, empowers new employees with intuitive search capabilities for a seamless onboarding experience.

Production Planner Analyst

Full-time

In close cooperation with the Brewer and Industrial Manager, I assumed various roles of responsibility within the industry, such as planning and supervising the production lines.

Industry:

I was responsible for creating Spreadsheets/Dashboards for planning the weekly production in collaboration with the Industrial Manager. I also helped line managers with controlling the press, reducing production losses and launching daily production.

Administrative:

Monthly closing of all raw materials and inputs used, carrying out the entire launch and conference in a ERP system.

Deploying Machine Learning Models

Performing deploys in different frameworks and tools for local or cloud consumption, such as AWS, GCP, Azure, TensorFlow, MLFlow, KubeFlow, Streamlit, Flask, Spark MLLib and Scikit-Learn.

Python for Data Analysis 3.0

The greater the number of tools we master, the greater our ability to solve different types of problems, isn't it? So why not learn to work with Data Science with one of the biggest and most powerful programming languages in the world. Well, that's what I decided to do.

Introduction to statistics

A course that introduces the basics of statistics and the uses.

Fundamentals of Software Engineering

I learned how to develop modular, documented, clean and efficient code, how to perform software tests, review, code versioning and how to generate and analyze logs through application debugging.

Introduction to Big Data

I understood how big data is impacting the modern world and how the data era is progressing more and more with new data and information... In the course, approaches and uses of big data and Apache Spark.

Big Data Fundamentals 3.0

I further reinforced the concepts learned in my previous big data course and learned new techniques and knowledge about data lake and data warehouse for data storage.

Google Analytics Advanced

BTech, Big Data Technology and Analytics

After an arduous 3.0 years, I achieved a bachelor's degree in Big Data and Analytical Intelligence from the Pontifical Catholic University of Paraná (PUCPR).