sambucini

yes, that's why i mentioned *for many domains which don't generate new data super quickly* ...

What troubles me about time series foundation models is that it's impossible to get a quick answer of whether these models work out of sample for many domains which don't generate new data super quickly. So, yeah, they gave an example of power price forecasting (my domain), so I would love to built a nice model with fancy covariates and all and backtest over the last couple of years -- but I have no way of knowing whether Mistral would be leaking data form its training into my backtest, no?

i always quite liked the flexibility of jira and the ability to logically connect tickets etc. I can see how it's perceived as this clumsy corporate tool, but i often whish gh issues had more of the features jira has.

I've been trying keeping an eye on open source issue trackers/project managament tools I can self-host -- with good cli/mcp capabilities. So quite happy to see this as I feel there isn't a lot! (currently also using gh issues) will check it out!

px dynamics GmbH | Senior System Engineer, Full Stack | Full-time | Berlin/Cologne/Hybrid | https://www.pxdynamics.com/

Background: px dynamics was founded by me less than half a year ago, I have 15 years experience in short term power trading in European markets, esp. algo trading. I do some consulting to bootstrap the company with and build customer relations. Simultaenously, we're starting to build a product business. For that we have strengthened the team with a Senior ML Engineer (ex google) and a Head of Data with prior experience in setting up data pipelines in this domain. The products we build are in a broad sense trading signals, e.g. to help traders of renewable or flexible assets trade their positions better.

Role: So given that it is pretty early days, we're of course looking for a broader skillset to lead the development especially of the cloud and backend infrastructure (hetzner), but eventually also frontend applications. Techstack so far is obviously minimal (linux, python, bash, postgres/timescale), so there is plenty of room to also shape this. What we as a team value is simplicity, really clean code/architectures, excellent documentations and quality over speed. So we're not trying to glue together the next venture to scam a bunch of VCs, but to really build something excellent.

if that resonates with you, please reach out to me directly via the email address you'll find on our website.

excellent -- thanks a lot!

I find this super interesting!! while having a quick look, I got the impression, that Mercury runs 1 web service for 1 notebook right? For me it would be very helpful to something like a dashboard with 1 webservice supporting n notebooks.

I'm using Papermill to operationalize Notebooks (https://github.com/nteract/papermill), it e.g. also has airflow support. I'm really happy with papermill for automatic notebook execution, in my field it's nice that we can go very quickly from analysis to operations -- while having super transparent "logging" in the executed notebooks.

thanks for that! we're all on-premise with our infrastructure, so I probably won't be allowed to use AWS here (I could imagine that the aws parameter store is also even more helpful in an architecture with several aws components)

i wasnt aware of the versioning bit in couchdb -- will look into that! thanks a lot!

thanks, looking into it!

yo, i work in the energy industry. it's not quite as terrifying as it looks. energy retailers typically sell electricity to end users at a fixed price. prudent retailers _hedge_ this exposure, i.e. they buy the electricity on wholesale markets that they promised to delivery to you at a fixed price, in order to reduce their risk to wholesale prices going up. wholesale price went up (A LOT) which drove many retailers across europe into bankruptcy, but well - they didnt hedge so well.. if you want a proxy for retail electricty prices is worth checking the wholesale markets, you can do that yourself on EEX's website: https://www.eex.com/en/market-data/power/futures

you'll see the following prices for trading day 2021-12-29: Year 2022 settled at 225EUR/MWh or 22.5c/Kwh -- which is admittedly crazy high. But year 2023 settled at 128 EUR and the following years below 100 EUR. This is plausible as we the high prices we have now provide a strong incentive to build new electricity generating facilities in the future. and wind and pv are already profitable at prices 50-80 EUR/MWh without extra subsidies. lastly: there are still enough retailers selling you power at prices below what your retailer wants to charge you, check https://verivox.de/ or the likes for alternatives

i've used prophet along many other ts methods for price forecasting in energy trading. my experience is that prophet is ok, but rather opaque. having tried many packages, I've always come back to the classical statistical methods seeing benefits in transparency (what's actually going on? impact of regressors?), speed and most importantly that these methods force the user to think about what's happening in the data and make conscious decisions about how to model things. but i can see that sometimes you dont care too much about accuracy and understanding but that you just want a forecast for something that works decently without much hassle.

yeah, it's both what you say and sales people's poor understanding of how annoying that is (and the fact that they get rewarded for closed contracts where operational costs dont matter)

I work in energy trading in Europe and for us this is a blessing. Of course we store our time series TZ aware, but many less professional small utilities or other market participants (incl. Grid operators) who e.g. Order power on hourly resolution with tz-unaware excel spreadsheets via email are a nightmare as each participants treatment of double or missing hours is different and requires manual processing.

I wasn't actually paying. The account basically dormant and from back when they started and courses were free. I am just trying to minimize my digital footprint by deleting accounts I don't actually use (I'm happy with value for money at coursera).

fyi - given it's annoying to collect all that information, I felt udacity might actually be able to help me with that. So I filed a subject access request under GDPR and asked them to provide me with all the data they have in relation to me..

Retain with pain..

The other day I tried to delete my udacity account. No button in settings, I mailed support, they asked my to mail legal (apparently they don't fwd mail..). From legal I received the following response:

Hello,

We are in receipt of your request to delete your account. We are sorry you want to leave the Udacity family; you will be missed.

Before we can proceed with fully processing your request, we need to verify your information. We take these steps to minimize risk to the security of your information and of fraudulent information and removal requests. Specifically, we ask that you provide the following pieces of information:

- Username and email address associated with your Udacity User Account (if different from the one you provided in your initial request);

- Online Courses currently or previously enrolled in;

- Approximate date of User Account registration;

- Country of residence; and

- A statement under penalty of perjury that all information in your request is truthful and that this is your User Account or that you have the authorization to make the request on behalf of the owner of the User Account.

If, after you have provided the above information, we are unable to verify your identity and/or authority to issue the request, we may reach out to you for further verification information.

As a reminder, any deletion actions we take in response to your request are not reversible and may result in Udacity (or you) being unable to retrieve information about your account, enrollment, and records of completion. Please also keep in mind that all removals of such information are subject to requirements to maintain certain data in our archives for legal or legitimate business purposes.

If you have any questions about this request please see our Privacy Policy or let us know.

Thanks for your understanding.

Udacity Legal Team

Great, thanks all!

Also Lithium ion battery have a fairly fixed ratio of capacity and power, of about 1MW to 1MWh. That makes them expensive for storing large amounts of power. With pump storage you're much more flexible, just build bigger reservoirs. very large pump storage assets can be optimized even seasonally.