develop, train and deploy a Machine Learning model

It is necessary to develop,train and deploy a Machine Learning model which will pick specific
elements from input data vectors. An input data is represented by a vector of 1920 elements of
floating-point numbers saved as NPY array. Example input data is shown on a graph, where red
lines represent indices of points of interest. An output data must be represented by a NPY
vector containing indices of found points of interest.

Training data will consist of input NPY arrays, pictures of input data graphs with marked points
of interest and “csv” files containing indices of points of interest (see attached data)


After given data analysis, answer please following questions:

Q1 - What type of ML model is capable to do the task, in your opinion?
Q2 - How fast the model will be? (We have around 2-3 seconds to process 107 pieces of data)
Q3 - Can this model be run locally, without active Internet connection?
Q4 - Approximately, how many pieces of data is necessary to teach accurate model? Right now we have around 8800 pieces of data, with only 60 pieces labeled. We are planning to start labeling the rest as soon as somebody will tell us that it is possible to accomplish the task.