How to make (almost) anything

G protein-coupled receptors (GPCRs), also known as seven-transmembrane domain receptors, constitute a large protein family of receptors that sense molecules outside the cell and activate inside signal transduction pathways and, ultimately, cellular responses.

My project would be to design a large scale sensing apparatus that will be based on the function of GPCR's.

This device will be an immitation of a human cell expressing a kind of GPCR's in its membrane. The nucleus of the cell will act as a server and each G-protein as a client, as far as communication is concerned. The receptor proteins will take input from the invironment, do local calculations and then broadcast the data through the G-Protein.

It is all about making... neuron cells! The goal is to create an ensemble of discrete devices morphologically like neuron cells, that when connected will act as a whole. Every single device will have three main parts: the dentrites (post synaptic terminals), the soma (processor) and the axon (pre synaptic terminal). The information from one device to another will be propagated by touch, through electric means. Each neuron will acquire the information from his neighbours and propagate it to other neurons through it's axon. Every given conformation of the ensemble will create a different organization of a "neural network" that can be trained to perform computational tasks.

The neurons will be organized like the ones in the human visual system. There will be a first layer of primitive sensors, like our rhodopsins in the retina. Next, 2 or 3 layers will form simple recognition systems for lines, shapes and colors. Finally, all the outputs of the intermediate networks will be fed to the central processor which will perform the final recognition/computation.
Unlike artificial neural networks, this network will have the neuromorphology of the human brain. The biggest difference is that in every node of the network, if you sample the output you will obtain meaningful information about the input, for example the color or the shape of it. In contrary, node outputs in the hidden layers of artificial neural networks reveal no information about the input.

Dielectric spectroscopy (sometimes called impedance spectroscopy), and also known as electrochemical impedance spectroscopy (EIS), measures the dielectric properties of a medium as a function of frequency. It is based on the interaction of an external field with the electric dipole moment of the sample, often expressed by permittivity/impedance.

Currently, spectroscopy in the regime of GHz is performed using network analyzers which cost thousands of dollars. I will try to implement a device that will cost a few hundrends of dollars, implementing a very fast sample and hold circuit pairing an ADC converter.