Skip to content
View lfvvercosa's full-sized avatar

Highlights

  • Pro

Block or report lfvvercosa

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
lfvvercosa/README.md

Luiz Verçosa

Senior Data Scientist and ML Engineer, Ph.D. in Computer Engineering. I build end-to-end AI systems — from classical supervised and unsupervised models to LLM applications serving real users in production.

Right now I work on a GenAI email-processing pipeline for the Supply, Trading & Shipping division at bp, and I do postdoctoral research on machine learning and process mining at the University of Pernambuco.

What I work with

  • GenAI / LLMs: RAG, LangChain, LangGraph, prompt engineering, Azure OpenAI, Amazon Bedrock, Gemini
  • Machine Learning: XGBoost, LightGBM, neural networks, LSTM/GRU, time series, clustering, SHAP
  • Platform: Python, SQL, AWS, Azure, Databricks, Spark, Docker, Kubernetes, MLflow

About this profile

Most of what I build is proprietary — the pipeline at bp, a RAG system in production at a Brazilian state court, a credit scoring product — so it isn't here. What is public is mostly research code from my Ph.D. and standalone examples.

A few repositories

nano_gpt A GPT built from scratch in PyTorch, step by step from bigram to full Transformer, with a head-to-head comparison of what each architectural piece actually buys you
example_playwright_lambda Running Playwright inside AWS Lambda for serverless scraping, container and IAM setup included
brazilian-justice Process mining over the Brazilian Justice event log, from a dataset I published on 4TU and Kaggle
discover_analytics_analysis_lawsuit Code for a paper on predicting lawsuit duration with machine learning and process mining
HeuristicMiner A from-scratch implementation of the Heuristic Miner algorithm

Research

15+ peer-reviewed papers on machine learning and process mining — Google Scholar. Ph.D. thesis on clustering methods for the performance analysis of legal processes, with a research period at RWTH Aachen University.

Contact

LinkedIn · lfv.vercosa@gmail.com

Pinned Loading

  1. example_playwright_lambda example_playwright_lambda Public

    Serverless web scraping with Playwright on AWS Lambda, containerized deploy with IAM setup, ECR push and invocation scripts.

    Shell 5 3

  2. brazilian-justice brazilian-justice Public

    Process mining on Brazilian judiciary case data — builds XES event logs from the public dataset and runs the analysis pipeline behind my Ph.D. research.

    Python 2 1

  3. discover_analytics_analysis_lawsuit discover_analytics_analysis_lawsuit Public

    Code used for the analysis present in the paper entitled "Investigation of Lawsuit Process Duration: A Machine Learning and Process Mining Approach" submited to the journal Discover Analytics

    Python 1 1

  4. HeuristicMiner HeuristicMiner Public

    A simple implementation of Heuristic Miner Algorithm

    Python 1

  5. analise_processos_eleitorais analise_processos_eleitorais Public

    Process mining analysis of case duration in Brazilian electoral courts, code for a published paper on procedural delays in the judiciary.

    Python

  6. nano_gpt nano_gpt Public

    Character-level GPT built from scratch in PyTorch, step by step, from a bigram baseline to a full Transformer. Following Karpathy's "Let's build GPT" lecture.

    Python