JULIUS MWANGI's Projects
The primary goal of this case study is to analyze data, identify patterns, and propose informed, data-driven recommendations that governments and stakeholders can implement to effectively address and reduce unemployment rates, particularly focusing on the African context
StarterNotebook
Fast data exploration using Sweetviz
⚡ boost inference speed of T5 models by 5x & reduce the model size by 3x.
Intended for an AMLD 2024 workshop.
Forecasting Time-Series Data with Facebook Prophet, published by Packt
Current available solutions eg #WorldCover apply inconsistent "cropland" definitions from #FAOSATA’s and are rarely up to date. This work focused on building cost effective classification for cropland mapping for #Sudan, #Iran & #Afghanistan regions. For Afghanistan we focused on temporal classification while for the rest year-long classification
This problem statement focused on building machine learning models that would assist create city-level air pollution susceptibility maps with a 5-meter spatial resolution for milan city in Italy. This city has unique challenges of dealing with pollution levels due to its unique topographic features.
The traditional in-situ soil analysis methods are laborious & inefficient, limiting scalability and hindering timely access to crucial soil data for optimal fertilization by farmers. In the amazing challenge, we tried to predict soil parameters(Phosphorous, Potassium, Magnesium and pH)from hyperspectral satellite images.
The fifth-generation(5G)of radio technology, is driving a profound transformations in connectivity eg faster speeds & lower latency. However, it introduces a significant challenge: energy consumption. There is thus need for accurate energy consumption modelling to save the environment and save on costs for telecoms.
This problem statement involved predicting fault impacts on Radio Access Networks KPIs and prioritizing issues that affect data rates—a critical step toward enhancing network performance and preventing customer churn in the telecommunication space.
Course to get into Large Language Models (LLMs) with roadmaps and Colab notebooks.
Build a deep neural network with keras that functions as part of an end-to-end machine translation pipeline
Mastering NLP from Foundations to LLMs, Published by Packt
Predicting Forex Future Price with Machine Learning
Repo to supplement my tutorial on Monte Carlo Simulations and Importance Sampling
This is a web service that allows people with medical Issues describe them in Swahili using voice. The solution will transcribe voice, send transcription to LLM. The results are not medical diagnosis but guidelines on how to mitigate their issues, over the counter drugs they can use before seeing a doctor. Users can listen and download the output
Neural Machine Translation (NMT) tutorial. Data preprocessing, model training, evaluation, and deployment.
Preconfigured Mt4 & Python Trading Enviroment
Useful Python scripts to help you be more productive
End-to-End solution connecting Metatrader4 & Metatrader5 💹 with Python with a simple drag and drop EA. Fully tested bug free & efficient solution for live & paper trading⭐ Full Documentation ready.