Were you Senior? I look at those situations as eat what you kill. Like I would ask my manager to put me in touch with arch or up level engineers to tell me where the focus is and find tickets to work on myself.
Yeah. I kept telling my manager and other team members I wanted to do some work. I also was clear that I didn't understand a lot of the tooling and systems primarily because I wasn't interacting with them, because I wasn't assigned any work (full circle). Over months nothing really came of this until at about 6 months later. However I participated in all the team meetings like scrum poker estimating tickets I had no idea about because I hadn't interacted with systems or software. Whether I was responsible for my own situation is debatable, but you bet if I've been idle for 6 months and not really getting adequate training that I'm definitely going to start looking for a new job. I've worked at companies of all sizes and never had this situation before and after, and at the time I was in Amazon I had been in the industry for more than a decade.
I'm not a webdev, but I suspect an overwhelming majority of their traffic is on mobile devices. So that's where a majority of eng time is probably spent. Not that it shouldn't be fixed.
Why not just use http retry-after? then you can use middleware/proxy to control this behavior. Downside here is that system operation becomes more opauqe and fragmented across systems.
I remember the original post the author is referring to. I was captivated by it and thought it was cool. When I ran the original mentioned in the post, it detected my one of my alt's that I forgot about. OP's newer implementation using different methodologies did not detect the alt. For reference, the alt was created in 2010 and the last post was in 2012. Perhaps my writing style has changed?
I usually just create a new account every time I get a new computer or reinstall OS. I thought most of the results here were noise, but after closer inspection it just found 10 accounts I forgot having. Actually incredible and a little scary how well it works.
It depends on your starting point. A baseline level of ML is needed. Otherwise ML platforms account for three basic functions: features/data, model training, and model hosting.
So do an end-to-end project where you:
- start from a CSV dataset, with the goal of predicting some output column. A classic example is predicting whether a household's income is >$50K or not from census information.
- transform/clean the data in a jupyter notebook and engineer features for input into a model. Export the features to disk into a format suitable for training.
- train a simple linear model using a chosen framework: a regressor if you're predicting a numerical field, a classifier if its categorical.
- iterate on model evaluation metrics through more feature engineering, scoring the model on unseen data to see its actual performance.
- export the model in such a way it can be loaded or hosted. The format largely depends on the framework.
- construct a docker container that exposes the model over HTTP and a handler for receiving prediction requests and transforming them for input into the model, and a client that sends requests to that model.
That'll basically get an entire end-to-end run the entire MLE lifecycle. Every other part of development is a series of concentric loop between these steps, scaled out to ridiculous scale in several dimensions: number of features, size of dataset, steps in a data/feature processing pipeline to generate training datasets, model architecture and hyperparameters, latency/availability requirements for model servers...
For bonus points:
- track metrics and artifacts using a local mlflow deployment.
- compare performance for different models.
- examine feature importance to remove unnecessary (or net-negative) features.
- use a NN model and train on GPU. Use profiling tools (depends on the framework) and Nvidia NSight to examine performance. Optimize.
- host a big model on GPU. Profile and optimize.
IMO: the biggest missing piece for ML systems/platform engineers is how to feed GPUs. If you can right-size workloads and feed a GPU with MLE workloads you'll get hired. MLE workloads vary wildly (ratio of data volume in vs. compute; size of model; balancing CPU compute for feature processing with GPU compute for model training). We're all working under massive GPU scarcity.
For the majority of usecases I have seen: solving a sufficiently large painpoint, understanding/formulating the problem, having/getting the right data, fitting well into a workflow of the users.
All the technology challenges are actually on the "cost" side of the equation. Meaning, that the aim wrt business value should be do as little of it as possible (but not less!). For some use cases this can still be quite a lot... But more often on the "all the pieces need to be in place for the whole to work at all" rather than "each piece needs to be super optimized".
this is really helpful, thanks. how much are third-party models changing these workflows (LLMs etc)? would you still spend as much time on feature engineering and evaluation? I'm wondering whether any saved time would be refocused on hosting, especially optimizing GPU utilization
I quit my job as a dev to care for my father who has dementia. Using COBRA and paying for the identical health insurance I had from my employer is $1,000/mo. I have no dependents. California minimum wage annual income is ~$32k. Almost half the income of someone who works a shit job. I'd say that almost 50% of minimum wage just for the privilege of going to a doctor when I have a problem is rightfully a pretty big deal.
This is the issue with American healthcare - it’s super confusing and people don’t even understand that they have better options. And of course, living with a super pricy COBRA plan while being unemployed is not the normal experience for people, and also not the actual “optimal” way to get healthcare in that situation.
COBRA isn’t meant to be a full healthcare insurance, it’s meant to be a “bridge” care for people in between jobs (that’s why it’s tied to your prior employers plans). You also don’t even need to pay the COBRA premiums unless you actually use it, so you can save that $1K/mo while being implicitly insured (helpful in case you only really want to be insured against catastrophic accidents).
If you were not working because you were caring for a dependent (like a sick adult), or if you’re a minimum wage worker, you wouldn’t use an employer sponsored plan through COBRA, you’d use Medi-Cal (California’s expanded Medicaid) or the ACA marketplace, and could be “free” or ultra low (premiums). Medi-Cal is free, and the ACA plan for a 45yo male in SF making Cali minimum wage (34k/yr) would pay $18/mo.
Healthcare is absolutely a “pretty big deal” and absolutely the system is terrible. But if you actually made 34k/yr, it probably makes sense to spend 1hr googling for the actual programs that exist so you’d discover that you can get it for much cheaper.
I take a biologic that costs $5,000 a month without a prescription. $1k a month is honestly easier and I can afford it. I enjoy the luxury of being able to afford to not have to navigate the stupid “oh, so you’re poor” patchwork bullshit that exists. My roommate is not such a person.
No, that's not how it works. Poor people who lose coverage from employer-sponsored health plans don't pay COBRA premiums out of pocket. They spend much less to either buy a subsidized health plan on the state exchange or go on Medicaid. The system is a mess but let's not exaggerate the problems.
I can’t imagine getting Medicaid to cover the biologic that I take that costs $5,000 a month without a prescription to be anything other than a nightmare.
Stack overflow was an invaluable resource for me as a developer. I got a wealth of knowledge about a bunch of thorny problems I was encountering with out paying anything. The people who answered the questions I had were able to help far more people than they ever could have individually without SO. I’m having a hard time being upset about his financial windfall. It’s really only a bait and switch for people who answered questions with some other unstated expectation than helping others or receiving help.
In aggregate, Stack Overflow probably saved developers far more time than anyone ever put in, and saved companies all over the world far more money than the founders made in their exit.
The founders monetized a platform where unpaid contributors built 99% of the value by answering questions for intangible "Reputation". The fact that contributions were voluntary doesn’t negate the issue. It highlights it.
A system designed to extract free labor for massive financial gain, while offering only gamified rewards, reflects exploitative design, not fairness.
This is such an incomprehensible take. Every developer has benefited from SO. I’ve contributed to the site exactly because it’s been so good to me. I don’t need money on top of that. The website is the reward. Nobody cares about reputation.
Websites cost money to run, and I’m glad SO was so well monetized that it didn’t run out of money and get deactivated like so many of the old school forums we used to go to for help before.
See my analogy below in this thread about the public park...
What is incomprehensible is developers confusing the value created by the site, with the means of production of said value.
While contributing to the site might provide you with some personal feelings of value, maybe even a rush of altruistic euphoria... :-) An illusory sense of digital sainthood...Your unpaid work directly enhances the value of a platform that profits from selling the collective output, including yours.
This means you’re effectively donating your expertise to a for-profit entity, with no equity or share in its success
Sure, they made money off my work, I don’t mind that. That’s fine.
Your analogy about the public park is wrong btw. A better analogy is if someone ran a private park for years where musicians can come and collaborate with each other and expertly exchange ideas, and then after 20 years of running the park they sell it and move on with their life.
Many sports clubs work with the help of many volunteers. If they were not available to help for free, the practitioners of the sports would need to pay more. That would mean, only the richer part of the candidates could do the sport. A side effect is as well, that the club itself gets bigger, more famous, and richer as a whole. All thanks to the help of volunteers. There is just no way around that.
In SO/Jeff's case this was beneficial for him in the form of a huge selling price. I would say luckily he was, as I read in the comments, a great contributor to the SO organisation.
Are there any command line tools that will help me with the drudgery of diving into an existing Java project using gradle and say something like “please update the dependencies of this package to this version using the existing project conventions in the entire project”? And give a changeset I can start validating and reviewing?
Most of my experiments have been IntelliJ plugins. I like what they have to offer but feel like they are quite far away from the unified understanding of the specific very complicated codebase I have to work in.
His first name is Brian. That’s his picture at the top. I can’t think of any other groups or
organizations that have the persona of a single person. Can anyone point to an example? Genuinely curious about this.
I always thought Krebbs was a cybersecurity firm organized like a lawyer's or dentist's office, where there is one senior person on the cover but they are rarely involved with individual pieces of work. Crazy to learn it is just one person actually, they do a lot of good work.
Happens in some artistic fields. Rodin didn't personally sculpt all of his sculptures. He directed the effort, but it was too much work for one person. I've seen Tom Clancy novels continue getting published even though he died over a decade ago. I think there are living authors doing the same thing, farming out production to ghost writers and just signing their name to the end product.
There are famous examples in advice columns, sort of. I don't know that any of them have ever been written by different people at the same time, but they've maintained stable personas and names even as the writers have moved on or died. The original founder of the Dear Abby column was famously the twin sister of the second iteration of Ann Landers and they feuded for the rest of their lives over it. They're both dead now but the columns go on using the same byline name.
I'm not parent, but at some point McAffee could refer to either the person or the company in the past.
Whenever I read a Wolfram blog post that floats to the HN frontpage, I'm never certain if the post is entirely the effort of just Stephen Wolfram, or is a group effort.
There's multiple interviews from over the years with people that worked very closely with him, including the person who was his publicist for a long time (or a very similar role, I forget exactly). You can obviously decide that it's all part of some long game of deception but it's just not needed. No one is trying to arrest him here and they haven't seriously wanted to for over a decade, he's become an institution like baked beans.
Well, for one, he can't film himself installing, e.g., fake artworks in the greatest museums in the world, as he did.
And, for those who don't know, he is from Bristol, England, the home of the band Massive Attack. I've been digging their music lately, especially their songs with the late Sinead O'Connor.
Using that name specifically has a bit of a different connotation than a generic one with no previous association like "Brian Krebs", though. If anything, it would be _more_ surprising to find out that someone going by the name Tyler Durden was just a single, regular person rather than something else going on.
