Much of my work from the past year is subject to NDA

This is a mix of directed and independent projects, and nowhere close to comprehensive. Of course, most of the complete compilable code was for coursework. But I'm most proud of the side projects that I haven't yet had time (there's never enough time) to finish.

A webapp in React.js with a Django backend. It's something like tweetdeck for telegram, and I'm building it for some open-source intelligence folks.

A webapp in Django that I did with a small team. Because of time constraints I ended up writing ~3/4 of this, and I think we all learned a lot. It's an expense tracking app, with three user types, and I used plotly for a graph of historical data, a fitted curve, and a prediction (the latter I did not implement single-handedly).

Ever wonder what your local climate will look like in a few decades? Me too. So I'm working on a webapp that predicts several climate variables for any county or state in the continental US (I know, but you have to start somewhere) as well as their 1st & 2nd derivatives. One piece of this is a database, and that's done, here is the code to build it (be warned, it takes quite a while).

A contribution to the rosetta code project. It's about Zumkeller (if you don't know this name, you should!) numbers, which are integers whose divisors can be partitioned into two disjoint sets each having the same sum. C++.

And here you can find a short little dynamic programming solution to a problem which keeps exactly no one up at night (except folks in algorithms & data structures, perhaps!). C++.

Implementation of the chinese remainder theorem to solve a particulary obscure problem (written in Java). This is unfinished because I thought of a simpler solution (if I had a dollar for every time that happened, I would... well, I'd have more dollars).

This is an ML quant finance project in python I worked on in the quantopian framework. I actually have a handful of these up and running, but this is a decent example. It basically picks the most predictive signals for a given symbol and then places confidence-weighted trades based on those inputs. Really, it is a signal processing problem, and very hard to do well. The code here is severely outdated but you can get the general idea. This certainly won't run outside quantopian.

What's where & why. Also, when. And how. And... you get the idea: maps are great.

Click chart to go see this fullscreen, otherwise it's trash.

Interactive version here. These images cut off the footer on export so the citations look wrong here. The bins are computed by pentacontile (yeah I just made that up. There's 50 bins in this map so it's not quintile binning...), so the color bar reveals a lot about the asymmetry of these data.

Notice the much more even distribution here.

Now compare those distributions to this.

FL, NC and a few other southern states stand out with low ratios, so does Texas though it's harder to see because the outbreak has been much more localized there (it deserves its own map, for sure). Why? Well I know a statistician got fired in FL for refusing to suppress valid data. Go see her FL COVID Dashboard. I live in NC and I know for sure that at least one county is supressing data. So I'd say it's unlikely that these states are just doing a great job of patient care. This is just bad data. By the way, here's the distribution (done in python, not flourish). Notice that using kde gives a nice smooth curve but the mode is at zero, so it implies some points are negative. But they're not of course. Looks a bit gamma to me though, might be worth trying to fit.

Here's a different, rather unusual look at the same dataset. Definitely go look at the interactive version.

This one has to be viewed interactively here. You just have to look at all the states. That said, this is an interesting relationship and probably worth digging into a bit more.

Humor is how I get through tough times. Interesting relationship here, school lunches are a proxy for many things so I'm not surprised there's a bit of correlation, but the outliers are the real story. Down there at the bottom is Yakima county, WA. The outlier way up top is Menominee, WI. The y-axis is plotted on a log scale. Counties are color-coded by state. In the interactive version you can hover to see what's what.

Here's my favorite.

The fullscale .pdf is here. The file is very large; it's a poster board, and you probably have to download it to view. There's also an accompanying paper that I wrote for technical writing. It's quite long (I won't be offended if you don't read the whole thing) but I really enjoyed this project.

I like writing, though I have not had much time for it the past few years (engineering programs continue to live up to their reputation as being... intensive).

My favorite little project is this little handout on the scientific theory. Not because it took a lot of effort or that it's particularly good, but because my professor tried to give me a 0 on it -- he said it was too good to have been put together by a student. What a compliment. I think he grudgingly gave me a B in the end.

Here is a very long research paper on climate change resiliency in North Carolina (also linked above under maps). I got inspired to do this when I noticed that while there is extensive climate resiliency research on a regional-and-above scale, only very limited information is available at the local level, particularly in locales where conservative politics dominate. Unfortunately, it is exactly those sorts of communities that are and will continue to see some of the most severe consequences of climate change. To get climate projections at the county level was a bit of an ordeal in terms of data preprocessing, and I had to get rather creative to downscale regional-resolution climate model data without completely losing predictive power. It was good fun.

A research paper on Thorstein Veblen and development economics can be found here.

Another one, this time on ethics. I wrote this for a chemistry lab, of all things. As if there wasn't already plenty to do!