Beacon Biosignals is seeking talented DevOps/Julia folks to help us take the task of analyzing hundreds of terabytes of EEG data from “ouch that hurts, my poor RAM, oof” to “painless, my containers are one big happy family.” Full job post here: https://gist.github.com/jrevels/7d3dcf28e69a25512fac39c765d3f8b8
We also have several internship slots open for the Spring/Summer 2020. Beacon has an abundance of low-hanging-fruit projects where intrepid Julianauts (perhaps YOU, dear reader!) could deliver a major impact within a 3 month period - reach out for details.
Contact me at email@example.com if interested!
Is all your EEG processing closed source and in house or are you using something existing in the Julia ecosystem?
Most of our EEG-specific processing tools are all internal for now, but are built on top of common packages like Flux, DSP.jl, etc.
We’re big on making as much of our work open as possible, though! We upstream improvements/fixes/features to our dependencies where possible, and are planning on open-sourcing a few of our packages by the end of the quarter
It may be of interest to your group that I’m trying to standardize some of the syntax for interfacing with computational neuroscience by implementing well documented properties based off of the Brain Imaging Data Standard. I’ve been working on it off and on for a bit but finally got an initial PR up for others to comment on here https://github.com/JuliaNeuroscience/NeuroCore.jl/pull/1.
If you would like to collaborate at all I’d be very interested as I spend most of my time working on large data analysis in neuroscience.
Coincidentally, we’ve also been working on data in BIDS formatted directories, and have just opened the first registration PR for BIDSTools.jl which is like a lightweight version of pybids: it parses metadata from a BIDS directory into an in-memory format which is simple to query.
Because the BIDS conventions are rather useful for data layout of medical imagery (regardless of whether it’s neuro related), we’ve been quite flexible and unopinionated about metadata schema so far, though that could be layered on top. I’d also note that we’re working with relatively small datasets (mostly neuro and breast MR imaging + spectroscopy) so we haven’t worried much about performance yet (eg caching of metadata indices).
I noticed this and took a look at that. You may notice I’ve even created an issue asking if you would like to collaborate. NeuroCore is by no means meant to do all the things that the BIDS standard does. It’s intended to be the ImageCore of neuroscience in Julia. That is, it provides a common syntax to build on.
For sure; BIDSTools is not planned to do everything either. A bunch of composable packages which can be layered as required is ideal for our purposes