Side Projects

About My Code Side-Projects

Side projects are a major way I add to my skills.

In addition to the ones below that are mostly done, There's also this repository with side projects that are partially done or just ideas. There's also visual examples of a few of the completed side projects in the gallery page.

Putting them in a list is a bit of a self-hack to limit the sprawl and push things to completion. It serves as a good reminder what isn't completely done and helps to prioritize.

What types of code do I write?

I write machine-learning, data visualization, and web development code. Mostly, I write in Python or JavaScript with smaller amounts of Ruby, R, PHP, sed/awk, C++, and Java. A lot of my side projects are in JavaScript as it is the easiest for me to jump into and quickly produce something with an easy to share visual.

Previous Side-Projects

• Assembled a Raspberry Shake, a personal seismometer.
• Built an internet connected pumpkin for Halloween that talked to small children.
• Presented a talk on the changing data visualization landscape in large organizations (1974-2016).
• Competed in a machine-learning contest to predict well log facies put on by a geophysics journal.
• Created an augmented reality webpage / business card using AR.js, which leverages three.js, aframe.js, and ARtoolkit.
• Used SVM machine-learning approach to identify direct returns, reflections, multiples, and coherent noise in seismic gathers as part of a Geoscience Hackathon organized by Agile Scientific and Total. I participated virtually and the rest of my team was physically present in Paris.
• Helped build map applications to assist in the spreading and collection of accurate information about shelters and flooding post-Hurricane Harvey in collaboration with a large number of other volunteers via SketchCity, a civic tech organization.
• Built version zero of an application that can take in any google forms results csv files, pick the right charts, and create a data visualization such that clicking on an answer to a question filters the answers to every other question.
• Participating in a geoscience hackathon run by AgileScientific before the annual Society of Exploration Geophysicists Conference in Houston. Built a python-based machine-learning model to mimic geologists’ stratigraphic picks of the top of the McMurry Formation in Canada. Still working on this project as I get spare cycles.
• Building a “where science happens in Houston” map that leverages web-scraping and machine-learning.
• Played around with using three.js to make three-dimensional data visualizations from car-based lidar data.
• Did a quick bit of machine-learning in a jupyter notebook to answer the question, “software engineer or data scientist”?
• Built a well log file uploader & parser called “wellio.js” that converts LAS files into JSON files, so the data can be manipulated by JavaScript. This enables other applications to be built that load local LAS files and manipulate/visualize them in the browser without needing python, server, or a proprietary application.
• Wrote a bunch of small JavaScript data visualization experiments on ObservableHQ, which is a new way to write public, live, forkable JavaScript in a way that feels a lot like a more easily sharable Jupyter notebook Specifically, messing around with machine-learning, data parsing, and sonification of data in JavaScript, often with geology or NASA data.
• Presented work on using machine learning to predict well log tops at the annual AAPG conference in San Antonio in May of 2019. You can find the Predictatops project on github with links to the presentation.
• Participated in the 2 day Glasstire Datahack. Glasstire is an art website with 18 years of art event data in the city of Houston. I combined the art dataset with an older half-cleaned dataset of locations for companies advertising science jobs in Houston to create a visualization of the distribution of Art and Science in the city of Houston. I leveraged a random forest model to clean the science data. Datahack code repository. Old science city repository. Hackathon product.
• Playing around in ObservableHQ.com with the notion of an explorable explainable for basic stratigraphy concepts related to sea-level and shoreface depositional environments.
• Building Wellioviz.js a javascript library for building well log visualizations on the web. While wellio.js focuses on converting LAS 2.0 well logs to JSON, there wasn’t really anything, at least open-source, to visualize well logs in modern JavaScript. G3.js existed but was no longer being developed. Plus, it relies on the older d3.js version 3. Hoping to eventually build a web UI so users can load well logs and change their visualization without any code knowledge or use of a paid product. Wellioviz.js takes in all plotting data and style instructions via a JSON so it should be possible to eventually connect it to Python code as well.
• Participated in Transform2020, an online week long meeting put together by SWUNG (Software Underground). Transform2020 was a great example of what is possible when you design an event with a digital first perspective as opposed to replicating an in-person meeting via teleconference. Transform had hackathons/sprints, unsessions , tutorials, and lightning talks. I participated in the hackathon by leading a sprint around wellioviz, and I also gave this 5 minute lightning talk about advice I give Pythonistas starting off in JavaScript. The slides were, of course, built with JavaScript.
• Created repository “geoVec-playground” that takes a glove model trained on thousands of geoscience papers (with focus on soils) and represents the words in a 3D embedding space using google’s stand alone embedding projector as a way to explore the model and gain a better understanding of its performance. Also, just interested in how to convert pre-trained glove models to tensorflow word2vec style format, so I can reuse some tools across different word embeddings.
• Wrote medium blog post “Alternatives to Iris: Finding Drop-in Replacements for Overused Example Datasets” as part of a thought experiment part of the way towards the goal I really want to get to, which is programmatically finding datasets that could be used for specific workflows / end projects based on characteristics of the known datasets used in those workflows.
• Finally got around to try some generative art code via some quick notebooks on Observable here and here. Pleasantly surprised how easy it was after the first little learning hump, so will try more of this in future.
• Whenever I try to get on the water be it surfing, paddle boarding, fishing, or sailing in the ocean, bay, lake, or river I check a variety of different sources of information. There’s a folder on my phone with 10 different pages. Brainstorming what I could do to combine some of those into one page via iframes or data visualizations. The first draft is a single GitHubPages page in a repo called water-check-houston.
• Doing a bit of research into application of deep learning and image recognition to generating data products that describe wave properties and surfers.
• Was recognized as a “Featured Creator” on Observablehq.com, a website / tool for data visualization.
• Wrote this Medium post about an in progress side project to create an easily forkable repository template that visualizes the de-facto community of related code & developers that is sometimes described by Awesome lists.
• Finally attempted to organize my side projects a bit with a README that splits out what is active, stalled, or just in idea phase.
• Wrote a little draft Observable notebook that takes your location and uses text-to-speech web API and a geology API called Macrosoft to tell you verbally about the geology of the rocks where you are. This was prompted by a tweet by DynamicWebPaige wanting an audio book about local geology she could use on a train and a previous Observable notebook by Ananya Roy.
• Redoing my website. I've been using a WordPress based site for 6 year and trying to recreate it in Next.js & tailwind.css with deployment to Azure Static Web Apps.

Side-Projects over the last couple months

• LLM Assisted Geology Descriptions of Arbitrary Locations = LAGDAL. https://github.com/JustinGOSSES/LAGDAL with live site that might or might not still be up as its not free to run. https://app-lagdal.azurewebsites.net/