AI, Volcanoes, Agritech and Data Strategy. I know, I'm confused by the combination too.
Working with the lovely people at https://fdl.ai
šOxford, ricocheting through a DPhil (PhD)
Folks - even if your primary motivation is research I'm begging you to spend time structuring your code beyond a notebook. I'm talking modularity, tests, docs, etc.
It will feel square peg -> round hole to start but once you find that new groove it will actually help make your work even more relevant which is what almost all of us are after anyway.
@arynn Snowflake has definitely been the most headache-free storage solution I've used recently. Bigquery can be pretty nice to use too. Conversely, Synapse was probably the most painful, with so many under-the-hood missing features at launch.
@arynn There is a lot to unpack here - if you lean into #lean concepts, then every activity should be related back to business needs. Furthermore, the weak link here might be the dashboarding/visualisation/reporting, rather than the data exercises themselves. That being said, unless there are privacy, security, or major cost concerns, playing with data in a sandbox is a good way to develop the capability of a team, and you might just discover something! #data #sandbox #dataviz #analytics
"I hope this email finds you well"
How your email finds me:
RT @MatToddChem@twitter.com
40 fully funded PhD scholarships available at @UCL@twitter.com for students from any country via the new Research Excellence Scholarship scheme https://www.ucl.ac.uk/scholarships/research-excellence-scholarship Deadline Jan 13th. @DrZoeWaller@twitter.com @profgeraintrees@twitter.com @duncancraigucl@twitter.com #phdchat
š¦š: https://twitter.com/MatToddChem/status/1590069392208850944
I wish I had discovered the nb_conda_kernels library for managing conda kernels in Jupyter sooner. That being said, I wonder if there is a less clunky solution for automatically finding, adding, and removing kernels for the Jupyter interface?
Mastodon habits I'm trying to lock in, rather than revert to my Twitter habits:
1) use CWs liberally
2) when threading, set first post to "public" and the rest to "not listed"
3) don't forget the description text when posting images (had to work on that in Twitter too)
4) throw in hashtags like it was Tumblr or Instagram when you want to reach beyond your followers
5) pin and visit hashtags to find more people
6) boost a lot
I think I broke the thread :-(
It carries on here: https://mastodon.online/@duncanwp/109308827893221490
Rude of me not to introduce myself!
I'm a #climate scientist at the University of #Oxford trying to better understand the role of #aerosol, in particular through their interaction with #clouds. I'm increasingly using #machinelearning to help emulate and interpret the huge amount of #satellite and #model data we have.
Perhaps also relevant: #emissions #uncertainty #feedbacks, and occasionally #photography š Feel free to reach out!
@FractalEcho: Depends on whether you're looking for low or no-code, but Cartopy is a pretty solid simple tool - I think it span out from the UK Met Office iirc. Simple to build into an in-house tool, not sure about its scalability however.
@arynn: Very few cloud migrations I've seen or been involved with have ever moved to a pure cloud architecture, but rather to some form of hybrid solution. While getting the networking, APIs and security right is important, the best implementations are when the data architecture is split appropriately for the business needs - whether that is per component, per row or per attribute. #cloud #cloudmigration #hybridarchitecture
Seems like there is a memory leak issue with Python 3.11 (cpython).
@madhugandha: The only place where I see a good argument in separating the two is when we talk about "general", "strong" or "broad" Artificial Intelligence, which doesn't yet exist.
previously, there was a stigma over using "AI" as an overly elaborate description of the work done by researchers and analysts, so "ML" became a more accepted term. In addition, other uses of "AI" were popular, such as video game "AI" which muddied the waters. These days, its mostly a semantic legacy.
@madhugandha: The only certain answer I can give is that Data Science is a collection of analytical techniques and actions, and AI and ML are simply a subset of Data Science.
Re the AI vs ML argument: I know researchers who use them interchangably and those that argue they have separate definitions - largely, in the modern data science world, both increasingly tend to mean the same thing and most people I know will either use both, or pick and stick to one for simplicity. #ML #AI #DataScience
@alice_i_cecile This thread is fantastic for finding active key ML researchers https://mastodon.social/@mathieualain/109292623186944594
Hope it helps š
Hello world! #introduction
Iām Richard, and Iām currently a PhD researcher at #Oxford applying #AI and #MachineLearning to #Volcanology and #Seismology . A ādata first, analytics secondā advocate, Iām also a #DataVault2.0 Practitioner and speaker, occasional data consultant and ex #Agritech CTO. Always keen to chat ML, Data Strategy, Volcanoes or anything exciting in the Applied ML world.
You can find me knee deep in #Python, #PyTorch, a few flavours of #SQL, #Snowflake and #GCP
