JurisGPT: an AI-powered Summarization System for the Supreme Court Rulings of the Mendoza province, Argentina
The project aims to develop an MVP for a summarization system that utilizes the rulings of the Supreme Court of Justice. Users will have the capability to produce summaries based on the court's jurisprudence. The system employs several technologies such as LangChain, a local LLM Vicuna 7B, and Chroma DB, among others, and will be deployed on the AWS cloud. This comprehensive solution will enable efficient and accurate retrieval of information from the court's rulings, enhancing access to legal knowledge.
One of the main challenges of the project are:
- Processing documents in Spanish: The LLM models will be assessed based on their ability to process documents in Spanish effectively.
- Fine-tuning the LLM for Spanish legal vocabulary: A crucial task involves adapting the LLM to accurately handle the specific legal terminology in Spanish.
Link to the project: link
Implementing a simple Retrieval QA system using LLM Vinuca 7B locally in Spanish. Link
Technologies: Langchain, Huggingface, LLM, Chroma DB, Vicuna LLM 7B, Text Generation Web UI, LoRA Fine-tuning, Summarization, Retrieval QA, Python, Jupyter notebook.
In this Data Science Challenge, two task were addressed.
Understand the data and prepare the pipeline to transform the data from the raw format to the requested form.
The goal of this task is to solve business case involving the development of the analytical model using the provided data.
Link to the project: link
Link to the conclusion presentation: link
Technologies: Python, Jupyter notebook, Numpy, Pandas, Seaborn, Data Wrangling, Exploratory data analysis, Scikit Learn, Supervised learning, Random forest classifier, Linear regression, glnmet regression, Feature selection, Residual analysis.
How many images per class are needed to efectively train a YOLOv5 network to detect defects in metal nuts?
Link to the project: link.
Technologies: Python, Label Studio, YOLOv5, Object detection, Cross validation, Docker.
This project is a dashboard built using R Shiny that allows users to visualize and analyze species observations on a map. The main goals of this project are:
-
Visualize Species Observations: The dashboard provides a map that displays the locations of species observations. The observations are represented as circle markers on the map.
-
Search Species: Users can search for species by their vernacular name or scientific name. The dashboard provides an autocomplete field where users can enter the species name and select from the available options.
-
European Country Dataset: The dataset used in this project includes observations from several European countries. Data is taken from the Global Biodiversity Information Facility (https://www.gbif.org/).
-
CSS Styling: The dashboard has been styled using CSS to enhance the visual appeal and improve the user experience.
-
Deployment: The application has been deployed to shinyapps.io, making it accessible online.
Visit https://rodralez.shinyapps.io/biodiversity/ to access the application online.
Link to the project: link
Technologies: R, R Shiny, ShinyDashboards, Leaflet maps, CSS, JavaScript, Shinyapps.io.
The R Shiny application efficiently addresses the following tasks:
- Users can enter a gene symbol or a GO term in the input box.
- Autocomplete results are displayed as the user types.
- If a gene symbol is selected, the application shows its corresponding gene symbol, gene synonyms, Ensembl ID, and associated GO terms.
- If a GO term is selected, the application displays the genes associated with it. For each gene, it shows the gene symbol, gene synonyms, Ensembl ID, and relevant GO terms.
You have two options to run this application:
- Execute the
app.R
file using RStudio. - Visit https://rodralez.shinyapps.io/gene-app/ to access the application online.
Technologies: R, R Shiny, Shinyapps.io.
A probabilistic framework to solve how many packs have to be purchased to complete a Panini album.
Link to the project: link
Technologies: statistics, R, Markdown.
This is a shiny app based on rocker/shiny-verse.
Link to the project: link
Technologies: R, R Shiny, Docker.
Data Analysis of admitted students to high schools of the National University of Cuyo for the year 2022
Statistical analysis about the distribution of admitted new students to high schools of the National University of Cuyo (Spanish only).
Link to the project: link
Technologies: statistics, pdf scraping, R, Markdown.