flaxandteal.gitlab.io/talk-newtech-2020

Building sector agnostic
open source, on-demand, container-based,
geodata science platforms for the web

Why?

Infrastructure, not application

What other applications of the identical back-end?

Why open source, why open data?

Modelling

  • Go-Smart: Clinical Domain Model (EU-FP7: numa.ie)
  • FEniCS, Elmer, OpenFOAM
  • Key to de-coupling model & case
  • Abstract - Simulation - Case
    (pre-print: http://bit.ly/gs-cdm)

<simulationDefinition>
    <algorithms/>
    <parameters/>
    <numericalModel>
        <definition name="DFAMILY" file="[ DFILE ]"/>
        <needles>
    ...

Data Times: data.gov.ie + Google DNI

Lintol: Open Data Institute

Where are user-driven models useful?

Setting up researcher-driven & model trials

Providing localized algorithms

via a single SaaS

National/local government services

phil.weir@flaxandteal.co.uk

10 Challenges

& Considerations

with web analysis platforms

Security

  • 1: Sandbox
  • 2: Storage and transit
  • 3: Automate infrastructure

Flexibility

  • 4: Beyond blocks
  • 5: Power of Python/R/Julia/Rust...
  • 6: No intervention required

Lift & Lay

  • 7: Not tied to an IaaS

Reproducibility

  • 8: Track everything, for everybody's sake

Interfacing

  • 9: Can test locally, easily
  • 10: Simple, succint interaction libs

>>> from gosmart.parameters import R
>>> [r.idx for r in R.group("organ")]
[1, 2, 3]

Summary

  • Openness
  • Scalability
  • Security
  • Collaboration
Questions
(or data science, product dev, consultancy, training)

phil.weir@flaxandteal.co.uk // @flaxandteal

flaxandteal.gitlab.io/talk-newtech-2020