Projects

Comets are small, icy bodies left over from the process of building planets. They provide clues to our solar system’s formation and possibly even the origins of life on Earth. Despite being one of humanity’s oldest noted phenomena, comet data and imagery are scattered globally and not easily accessible to astronomers, impeding science.

COMA changes that. Aspiring to become the global resource for all comet imagery and photometry to help drive science on small solar system objects. Funded by NASA and built by the University of Hawaii’s Institute for Astronomy .

On the team, I’m primarily responsible for coding and architecture on our ingestion and backend database APIs. Our ingest pipelines process scientific imagery into a form astronomers can easily use and the APIs fuel querying the system by scientists as well as processing from our science backend and sources like NASA’s JPL. It also feeds our really nice frontend. Ingest pipelines and our datalake are orchestrated via Airflow , and our APIs are built on the excellent and highly performant Go Fiber framework.

I’m really enjoying working on this. Learning heaps both code- and astronomy-wise. Keen to do more like this and even more ambitious projects. So, if you found this via search and have interesting Astro problems and projects, or are interested in chatting with me about your astro/physics PhD program (I’m actively looking), mail me .

Habits make us. We become what we do regularly.

harsh is habit tracking for geeks. A minimalist, command line, open source app for tracking and forging habits.

After trying an exhaustive number of habit trackers that didn’t help build habits, I took the best ideas, distilled features, ruthlessly simplified, and provided easy ways to install (and onboard to) the app via homebrew , snap , and OS- and arch- specific binaries via an automated Github Actions release pipeline. It works everywhere.

There’s more about it on the release post . This was seriously fun. And marked Go as my go-to monkeywrench for heavy lifting and microservices.

A full 1/3 of the world’s ~8750 amphibian species are critically endangered or on the verge of extinction.

The AArk Species Assurance app allows zoos and conservation groups to coordinate rescue and science programs globally, allowing them to bring at-risk species back from the brink of extinction and combine resources effectively.

Amazingly, in production for over 9 years now, running tickety-boo with incredibly few problems thanks to the Padrino web framework and despite time and traffic. Major upgrades happened in 2017 and the very end of 2022.

The app and program is a joint partnership between WAZA and the IUCN SSC .

I loved working on this. Seriously keen to do more work on wildlife conservation projects. If you found this via search and have some interesting problems, email me .

DataKind is a datascience-for-good charity, helping NGOs answer difficult data science questions key to their effectiveness and helping them use and collect their data better.

When DataKind decided to expand to five global chapters, a trio of co-founders and I put forward Singapore as a site.

We won, started up, and I acted as Director of Operations for two years.

Kobayashi is named for the majordomo of the infamous Keyser Söze in The Usual Suspects and Star Trek’s no-win mission scenario (since blog software gets zero dev cred.).

Nothing in Rubyland did what I considered essential. So, I took the best ideas I’d seen, shmooshed them together, and created something super fast, hackable, easy-to-use, and with a tiny code base (<250 LOC!). I had a blast coding this and learned a surprising amount for what seems like such a simple project as I layered on more complex functionality.

A Leadnow project during the 2011 Canadian Federal election National Debate to allow a large constituency to discuss and understand details of policy arguments in the debate.

The app itself pulls tweets out of the twitter firehose for any arbitrary hashtag(s) (or any text in the tweetstream, for that matter) and dumps it into a database and then pulls it out slowly in a nice, ordered waterfall cascade-style for coherent discussion and response.

Leadnow contacted me and asked for help getting their fledgling operation off the ground tech-wise for Canada. The site was light, fast, and very scalable with an easily extensible codebase that was put up in record time and ran their online presence until late 2014.

My favourite thing about this entire project was that I did the initial work and pushed www.leadnow.ca live globally from a lovely little island in the Andaman Sea in Thailand while on vacation, digital nomad style at a time when that was very rare. I really need my life to be a bit more like that, more often. It definitely proved I could do remote work from anywhere despite geography and timezones.

Catherder came about since I work a lot with people whose big problem is that they work way too long and hard. Unhealthy overtime becomes the norm, burnout follows, and good people end up leaving organizations that can ill-afford the loss.

Tess allowed us to see trending counterproductive overtime, stick to legal TOIL (Time Off In Lieu) obligations, vacation, and sick days, and make sure staff stayed healthy, rested, and firing on all cylinders. Audit compliant, one-click approvals via email, nice reporting and alerts, this was an “off book” project with an easy weekend-to-PoC solution.

Catherder was further enhanced inside Pivotal Labs and rolled out for all APAC staff across 5 countries.

I still think its simplicity and ease-of-use for manager and staff make it far superior to most commercial HR systems I’ve seen.

A big problem for fast-moving activist organizations is they rarely have time or resources to focus on campaigning infrastructure until it starts to hurt core execution and actions

Brought in as Director of Online Campaigning and IT in 2009, our team redesigned and rearchitected the GetUp core campaigning engine in Rails for rapid response campaigning, action flexibility, email blasting at “Obama-esque” campaigning speeds, salient metrics and A/B testing, and an elastic cloud architecture designed to handle massive, yet intermittent “fireballing” of the site. As a benefit, using cloud-first technologies minimized costs during off-peak email blasts and TV media by backing off resources.

Lunchmeets was designed to solve a seemingly small, but big problem Amnesty had, people not really knowing the people they worked with. Collaboration and sharing are vastly more difficult with people you don’t know (and trust), and a showstopper if you’re trying to change an organizational “gatekeeper” culture. This app randomly matched people up to go to lunch or coffee.

Another “off book” project, hugely successful with users and had been running in production since mid 2007 to late 2013 with thousands of Lunchmeets gone-on and over a third of all full-time staff participating.

Amnesty International needed a complete overhaul of their web platform, online branding, and ability to reach and engage supporters in non-traditional locations globally.

A massive project involving overhauls of everything from underlying technologies, to branding, to content, to the document library at the heart of Amnesty’s human rights research, this comprehensive relaunch of the amnesty.org flagship site delivered capabilities the organization had never had before as well as a key deliverable for faster turnaround for digital campaigning and rapid response actions.

This was the very first Rails app I ever built, way back in 2006 when the framework was at 1.2.

Built over a sleepless weekend to meet the urgent need to get good information on voting patterns for the original federal leadership campaign of Elizabeth May of the Green Party of Canada . It was simple, tightly featured, and did precisely what was needed.

It gave the national campaign team daily statistics on canvassed ridings which was critical for understanding where to focus resources for maximum effectiveness to win the election.

I like to think this app helped contribute to May coming from behind in the polls to beat her opponent by a 2:1 margin. Elizabeth herself actually personally thanked me on national television with a really nice mention in her victory speech in Ottawa.

Knowledge management is a deep problem for all organizations. Vendors intentionally confuse buying software with creating a culture of knowledge sharing and findability.

Whuffie tried to solve the problem my new organization had of being great at writing things down (explicit knowledge), but with low organizational learning value. Tacit knowledge in peoples' heads had more value but was inaccessible, so Whuffie was an experiment to deal with questions and dialogues and socially reward people breaking down a “gatekeeper” culture by creating a marketplace for knowledge and help while
encouraging information sharing as something valued by the organization.

For scifi fans, the project takes its name from Cory Doctorow’s Down and Out in the Magic Kingdom where Whuffie, a form of post-scarcity digital reputation and ersatz currency, replaces money and is a proxy for societal value.

This ML system extended from my MSc work methodically sifting satellite imagery to try to find undiscovered archaeological sites.

The idea was inspired by NASA’s remote sensing help in the 1992 discovery of Iram of the Mighty Pillars , the famed Atlantis of the Sands in Arabia’s Empty Quarter (the Rub’ al Khali/الربع الخالي‎‎), via ground penetrating and Synthetic Aperture Radar (SAR) from satellites.

If you’ve run across this via search, I love applying technology to archaeology, so if you have an interesting project, please mail me .

I was maintainer on this from 1999 on after taking over from the original author, Keith Resar . The code automated virus definition updates across large corporate networks for McAfee AV and let the sysadmin know it had been done via automated mailouts. At peak, it had 45k+ installs worldwide and saved a number of corporate networks during at least two major global virus 0-day outbreaks which took out less prepared IT groups.