Community

Putting the Labour/Blair thing to one side and responding to the fraud expansion comment, I do see UK Fintech leaders in real-time fraud detection and complex fraud detection (tax, AML, fincrime, etc). I've worked for two such firms.

Complex financial crime is easier to detect than it used to be. Money launderers, organized criminals and tax dodgers have far fewer places to hide. Also, streaming real-time analytics, including model inference at scale in-memory and with distributed memory, with much-improved model deployments (types and agility to recalibrate) is powerful and increasingly democratized.

That said, I'd also note that fintech favourite, crypto, is the criminal's friend and I'd hope there's a responsible crypto strategy in the bag of tricks of the next UK Govt.

Very good. I'd not seen it! Quite agree. Cassandra is my old stomping ground!

Hey Ketharaman Swaminathan. Thanks for the comments! Great to hear from you.

This dates me, but my baseball knowledge is centered on Randy Johnson and Ken Griffey Junior (apparently there's a junior version of junior I saw the other day). 

Ex.3 - having just watched a webinar earlier today on the failure of stress testing models (admittedly capital/risk ones rather than bog standard portfolios) with GFC and more recently Covid, I wonder if I should have waited before posting it because I might have added extra layers. My inspiration at the time was less the model itself but more around the democratization of the human linguistic prompt that any non-Quant/coding PM might make. After today's webinar, though, I might be tempted to throw in the real extreme scenarios into the mix, the black swan scenarios which the LLM might be able to make sense of, but my model perhaps can't "imagine". Thus the LLM output could adapt weights and correlate up dependent factors, possibly make sense of related factors that inform the "copula"-like thing which seeds the simulation, etc. Your point is completely valid!