Tiny Data Science projects by hour, with creative person with VERY STRONG statistics background (data scientist + Machine Learning)

I usually do "mini" Machine Learning projects for my students (2 or 3 per month).

My skills are not strong enough to develop some of these projects. So I'm looking for someone to help me to develop these small projects, when my other helpers are busy. It's important that they don't necessarily follow my particular approaches BUT choose and apply MANY different creative and innovative tools and approaches, in every project, because the idea is to show students different tools and techniques. All this jobs requires a VERY strong knowledge of statistics to test models, to calculate likehoods, confusion matrix , AUC, R2, etc with a very strong knowledge of statistics and probabilistic base . The projcts needs to be develop using Jupyter notebook in a easy to understand code (step by step) with basic documentation.

I don't like to receive a project and then have to go back and forth several times to finish it because it has bugs or some plots are missing. I like very much creativity and self initiative persons.

THIS PROJECT:
For this particular project is a "test" for a new freelancer to show experience creating/modifying an existing jupyter notebook ..(in this case it is a model to create a diet for population with previous learned recommendations" for a guarantee of success....).

I have already created a model of 48 categorical variables (already created and tested) (21 variables "before diet" and 21 "after diet". (the model includes some other features and an output variable that summarizes the success.)

My initial approach was to use Recomendation engines plus Machine Learning Techniques. Initially I found "similar" patients who were successful on their diet, I used Cosine similarity with Sklearn , which is already tested and working fibe but needs to be enhanced with plots and further analysis such % of predicted sucecess for similar patients, so and so)

Then, I also applied a machine learning (xgboost) to predict for new patients, with the best approach according to training, then I used SHAP , but should use different explainers (here i used the "before" diet feature set, to explain the importance of the characteristics for this particular intended case. ( how is the relative imporance of each feature ) .

The idea behind this, is to show students different models (recommendation engines and ML combined ) and then the new patient, will have his very own plan with all features explained.

I already have an "almost done" notebook for this project, but I need to look at different approaches. Please note that I'm not just looking for a one time project but these kind of "small" projects for the future.

I usually ask to apply K-means, random forest, xgboost, lightGBM, flask and sometimes projects includes SQL integration.

BUT most of the projects NEEDS to have extensively use of plot´s descriptive analysis.

Other projects uses spatial analysis with geopandas and bokeh as well.

Sometimes we need to webscrap some data (with Selenium and Beautifulsoap),

Sometimes also I ask for Computer vision and NLP projects

So, the strong skills to use these or similar tools is IMPORTANT and must be VERY robust.

PLEASE READ: the average time paid to develop each project is approximately 2 or 2.5 hours, (obviously sometimes more than that) but the datasets are usually ready to use. So the total amount of the project will be accordingly, remember that all these projects are not commercial but academic.

I will share my notebook with last three selected freelancers prior to select finalist freelancer, not with everyone.

I would like to see and understand your own approaches , how fast and creative you can be, your creative skills in plots and used models and art combination of tools and techniques to solve problems and produce an innovative solution to arrive to the desired final results.