JD Bothma
JD is Head of Product at [OpenUp](https://openup.org.za/) - an open data and civic tech non-profit.
A software developer by training, he supports work across the organisation to ensure there is a balance of focus on what clients, partners, and end-users want and need, and progress towards our organisational vision. That vision is a South Africa where citizens and government are empowered to thrive, collaboratively.
Having worked with massive legacy monoliths and many micro experiments, he finds it thrilling to build simple, robust systems, that reliably do their job, and don't do harm when they fail. Building things that make data and information available in ways that are meaningful to their users, and don't misrepresent data during partial failure conditions.
Building many data tools for users and administrators with a range of data skill, he's learnt a hard lesson or two about what it takes to make a data system maintainable and sustainable, but still has so much more to learn before this would be considered solved.
Accepted Talks:
Lessons learnt building and maintaining National Treasury's open budget data viz portal
We built an open data visualisation portal for National Treasury - https://vulekamali.gov.za/. Vulekamali had to make a range of structured and semi-structured data, and many other files, easily accessible to everyone in South Africa. And in ways that are meaningful both to experienced data analysts, as well as someone who never did high school accounting.
We built this using Django as a core component that manages and pulls together data from an open source Python data management system (CKAN) and an open fiscal data query platform (OpenSpending).
In this talk we share the architecture and implementation decisions that paid off, and the decisions we still regret, as well as our thoughts on how we could improve on them.
We'll cover things like building a system to to handle partial data, automating bulk file uploads, how hard it is to get people to format data correctly, and giving them helpful feedback automatically.
If you build systems where humans manage at least some of the data directly, you'll either learn something or teach us something in the comments.
