Jonas's Projects
A bot to hit the webpage of the Berliner Ausländerbehörde until you can get an appointment
This is an open source version of the CADI AI software.
Repo for analysing contributions of the private sector to disaster risk management.
Experimental Closed Domain QA Pipeline to build an application that allows to ask open questions to UN project documents.
Data in Climate Resilient Agriculture (DiCRA) is a collaborative digital public good which provides open access to key geospatial datasets pertinent to climate resilient agriculture. These datasets are curated and validated through collaborative efforts of hundreds of data scientists and citizen scientists across the world.
SDG AI Lab in partnership with UNDP DRT and CBi has developed an online tool – a Frontier Technology Radar for Disaster Risk Reduction (FTR4DRR), which allows for the systematic tracking and understanding of frontier technologies as they are developed. This would categorize technological solutions according to their technology type, disaster/crisis type and maturity level. Moreover, it is expected that the tool developed would encourage knowledge and experience-sharing among development stakeholders on the use of frontier technologies in disaster and conflict contexts. The Frontier Technology Radar for Disaster Risk Reduction (FTR4DRR) aims to highlight the potential of technological solutions in disaster contexts to those working in the fields of risk reduction, response and recovery. It supports development stakeholders to navigate the variety of existing and emerging technologies and their possible use cases.
Gives information on how to access High-Performance Computing infrastructure
A Telegram Mass Surveillance Bot in Python
Scripts to run large Meta NLLB models on DFKI GPU Cluster
Reference code base for ML Engineering, Manning Publications
Working Repo for building a set of models to automate the classification of project log-frames to a comprehensive taxonomy. Data can not be pushed yet.
My implementation of the original transformer model (Vaswani et al.). I've additionally included the playground.py file for visualizing otherwise seemingly hard concepts. Currently included IWSLT pretrained models.
Repo for experimentations and testing RAG (retrieval augmented generation) pipelines for analysing Ugandan audit documents
Machine Learning models to train on GPU cluster for classifying text according to the 17 Sustainable Development Goals
Development branch of Policy Tracking tool - Accepted submission to UN World Data Forum 2023
Template to create ML apps using streamlit and deploy it on heroku
Deploys trained ML Text Classification models and allows for user feedback to iterate and improve performance over time.
Analysis of all available UNDP Job postings over the last years