|
leper khan posted:I came up with that independently back in uni on an exam. Later courses added the word 'nondestructively' to the question. Step zero: fork()
|
# ? Jul 23, 2021 19:37 |
|
|
# ? May 23, 2024 20:23 |
|
give hash to linked list. store hashes in array. take fourier transform. look for peaks by sampling
|
# ? Jul 23, 2021 19:40 |
|
What are some good resources out there for former Java developers who wanna get back into the language but are a bit out of date? Say version 7 (2011) is the starting point. I use it at work but all more legacy apps. Got a decent opportunity coming up that I think would need me to step it a bit there. Worth learning Spring Boot? I remember regular Spring. Remember not liking it and the xml thingys.
|
# ? Jul 23, 2021 23:13 |
|
Comb Your Beard posted:What are some good resources out there for former Java developers who wanna get back into the language but are a bit out of date? Say version 7 (2011) is the starting point. I use it at work but all more legacy apps. Got a decent opportunity coming up that I think would need me to step it a bit there. Spring Boot is used everywhere these days, it's the default java framework, so it's worth learning and does not require XML.
|
# ? Jul 24, 2021 00:06 |
|
Comb Your Beard posted:What are some good resources out there for former Java developers who wanna get back into the language but are a bit out of date? Say version 7 (2011) is the starting point. I use it at work but all more legacy apps. Got a decent opportunity coming up that I think would need me to step it a bit there. You probably have more experience than me, but I find that there's always so many changes in each version, but most of them totally irrelevant to my use. As for things I found worth making note of was try-with-resources in Java 7, Lambdas in Java 8 and the var keyword for type inference in Java 10. I rewrote a Apache Camel project that used Spring to use Spring Boot and found it pretty straightforward even if you don't have a lot of experience with the former. I don't think it's going to come up again for me, but your experience might be different.
|
# ? Jul 24, 2021 01:56 |
|
I figured I would roll back to earlier in the conversation:Pollyanna posted:Very fair! This isn't do-or-die, I'm not coming into this expecting to succeed. I do wanna see how far I can get, though. Yeah I'm going through these motions despite in a position at my current job where my first two levels of management are pretty decent. The problem is that changing any of that ruins it and that stuff changes every other year, so I figured I'd go ahead. My impression of the FAANG interviews originally was that getting into this independent song of dance was a way to get into higher-level software places without having a decade of distributed web crap, so I gave this a lot of effort across 2019 and this year. I now don't think that's entirely true because I think my inexperience with distributed web comes up with the lack of common scenarios in my behavioral interviews, although I apparently actually pass my system design interviews (?). Anyways, this is about you and not me. Facebook felt like gatekeeping shibboleths in particular since they basically gave me 20 minutes a piece to do 2 medium+ problems, with an assumption that you could carry out dialog and thinking out loud at the same time. I could do that in 30 minutes and do so repeatedly in recent interviews, but 20 is really rough. When I first started doing these, I was absolute garbage at them and it took me awhile to get into the groove. I thought that cracking open algorithm books would be relevant, but they really weren't. I also thought for some reason doing some functional programming again would help and that was dumb. This year, I just focused on the coding problems. I did think getting a book with the problems and basic solutions for a ton of these types of problems would have been useful just to see what common stuff comes up, but I didn't really find one late in the game when it had occurred to me to try. I prefer Hackerrank to leetcode because Hackerrank problems tend to describe what makes the particular problem complicated. Leetcode instead just tells you that you're wrong despite and you kind of thrash. There were multiple questions where I looked into the discussion and found people complaining about alternate means to the same ends that would not pass despite conforming to the problem. On the other hand, I also did a few leetcode problems in order to practice dealing with assholes that can't explain their algorithm lottery questions well. It was somewhat helpful it telling if there was a "catch" to what seemed like a simple problem. With Hackerrank, I can reduce my "score" (who cares) in order to get a failing test case. It's usually some gajillion-mile input/output long but I can then dump it into a debugger and see if I can sort it out. On the flip side, my recent Google phone screen was 0% algorithm lottery and a normal object doing normal object stuff. It actually took me a few minutes to mentally switch my gears back into something normal. My final Amazon interview coding problem was a similar normal object thing, but with lots of snags and gotchas based on decisions I made. So you could work on all this poo poo not even use it on the screen, but perhaps for the interview. I started out coding into an IDE and then worked my way into Hackerrank's/leetcode's editor (or flat out using Notepad without any help). You should be fine throwing your hands up after an hour or two (depending on the problem) to see sample solutions. I liken it to learning marksmanship: after a certain number of shots, you're just kind of done for the day and doing any more may be counterproductive. Re: Dynamic programming: If the problem doesn't want to know the exact solution, like how exactly you got what you got, but just some number (best/worst/true/false) then it reeks of dynamic programming. Techniques you find online talk about finding the redundant paths and whatever, but I think the mental hangup is that many of the paths aren't literally redundant, but they just so happen to be coming to same outcome, which is what is "redundant." Alternately, once the DP alarm goes off in your head, you can assume you're going to make some intermediate data structure to store previous calculations. If you can't figure out any better way, then it might as well just be a dictionary. For all this effort, I'm kind of beholden on Google to consider an offer, because I washed out of Facebook's screen, Amazon gave me a zero-feedback rejection two days later, and now I'm just sitting on Google to respond from 1.5 weeks back.
|
# ? Jul 24, 2021 04:26 |
|
Anyone know a community that's good for ML questions like "how might one approach identifying cool accomplishments or recent praiseworthy things on someone's dating site or LinkedIn profile?" That's literally what I want to ask because it seems like fun as a first project in ML but then there are tons of tools out there, etc. Just from thinking about it, I kind of get the impression that some kind of supervised learning would probably be the right approach but then I start to wonder how I'd get enough data for that... And even before that, I don't even know how to estimate how much data would likely be needed. Maybe what I need is a book on ML that talks about these sorts of things from a more practical perspective rather than theoretical?
|
# ? Jul 24, 2021 16:33 |
|
and it'll also give you a crash course in not building a stanford-detector, woman-rejector, etc
|
# ? Jul 24, 2021 16:52 |
|
That's a much harder project than you think it is.
|
# ? Jul 24, 2021 17:32 |
|
ultrafilter posted:That's a much harder project than you think it is. What about it do you think makes it hard? If you had all the data you needed...* You'd need some way to read text and determine what's a subject. There must be someone who already trained a model for this and sells access to it as an API. I probably don't need many calls for a hobby project so shouldn't be expensive. Then I need to take those subjects and classify those into events and accomplishments. That seems like the harder part because I'd have to label them all somehow, meaning I'd have to think about what exactly counts as a notable event or accomplishment. But once I label enough examples, seems like I just need to pick some kind of ML library and apply a bunch of supervised learning techniques like linear regression or neural networks until I stumble upon one that seems to work? * I've heard getting the data is often the hardest part so maybe what I just described above is actually as simple as it sounds and the data itself is the hard part?
|
# ? Jul 24, 2021 17:45 |
|
just buy low, sell high - how hard could it be?
|
# ? Jul 24, 2021 17:48 |
One of the Ars Technica people just did a multi part feature about trying to use ML to assess headlines, might be worth it for you to read for inspiration.
|
|
# ? Jul 24, 2021 17:54 |
|
Osmosisch posted:One of the Ars Technica people just did a multi part feature about trying to use ML to assess headlines, might be worth it for you to read for inspiration. Was a good example of someone diving into a project in a subject in something they knew little about. Kind of gave me an idea of how a newbie like me could approach a hobby ML project. Started looking at small/beginner kaggle competitions and comparing some ML and deep learning courses. Will see if any of them see like they could lead to a solution. Kind of wonder if this is the sort of thing where I'd need to develop my own technique for creating a model rather than selecting some pre-existing model or even pre-existing technique.
|
# ? Jul 24, 2021 23:55 |
|
oliveoil posted:Was a good example of someone diving into a project in a subject in something they knew little about. Kind of gave me an idea of how a newbie like me could approach a hobby ML project. I'm the world's biggest noob, but Unity ML is a fun way to experiment with ML. You can train agents to play games or do little challenges. You can do pure reinforcement learning or the one where you record yourself playing and it tries to train an agent to replicate your style or you can do a mix. Might be worth a try
|
# ? Jul 25, 2021 03:42 |
|
if you wanna thing thats reasonably feasible to get done with actual value, find a dealio to classify with at least on the order of 10,000 datapoints (10^7, 10^8, 10^9 much preferred, interpolate accordingly) and then classify it. that's the happy fun easy land of ML agentic poo poo is the less-happy-fun-easy land and every tech major has like 3 or 4 deployed dealies tops outside of video game and other game poo poo, as opposed to literal hundreds of classification dealies for every tech major and prolly thousands at the goog. and of those 3 or 4 one is prolly a clone of the goog's RL dealie approach to minimizing server power costs (but in turn this is basically a regression with like two dealios attached) video game poo poo, go whole hog if you wanna, tho
|
# ? Jul 25, 2021 04:31 |
|
bob dobbs is dead posted:if you wanna thing thats reasonably feasible to get done with actual value, find a dealio to classify with at least on the order of 10,000 datapoints (10^7, 10^8, 10^9 much preferred, interpolate accordingly) and then classify it. that's the happy fun easy land of ML That's amazing. I thought I was late to ML but between this and looking at the kinds courses and books available and questions people ask... It seems like novel ML problems are only a thing for a subset of engineers so small it's virtually empty. Most people are just using variations of the same approach for everything but somehow there are job descriptions for "ML engineer" everywhere. It's bizarre. The complexity of existing jobs seems to be on the level of knowing which database to pick and how to configure it. I'd laugh if the average full stack dev job specifically wanted a "Database Engineer" just because you'd be touching a database at some point, but it seems like that's the level of understanding expected of people working in ML roles. Am I wrong? Literally just do some decent-sized classification projects and that's pretty much it? Doesn't seem much harder than setting up a few rails apps for my git repo as a student to prove I could actually build something and sketch out a normalized DB schema.
|
# ? Jul 25, 2021 17:48 |
|
There aren't really a lot of problems that require a moderate amount of machine learning knowledge. Problems for the most part are either solvable with off the shelf components or require some nontrivial amount of research. What's changed recently is that a lot of companies who previously didn't do any machine learning are looking to apply it to automate some of their processes. In a lot of cases, the problems that result really are easy enough to be handled by someone with basic knowledge. Unfortunately not every problem is, and companies are by and large not hiring people who can tell the difference. That's going to end poorly in at least a few cases.
|
# ? Jul 25, 2021 18:31 |
|
I can't believe I avoided this for so long because I thought it was pretty much entirely problems that require nontrivial research. Wow. Some of the research also looks like something you could figure out if you were good at puzzles and familiar with the material. Not most of it, granted, but some of it... Enough of it that I'm tempted to relearn linear algebra and maybe skim my old calculus or analysis books or something and see if there's any low-hanging fruit out there.
|
# ? Jul 25, 2021 18:51 |
|
Pollyanna posted:Lord help me, Google's interested, and now I'm staring at a pile of prep materials. My advice is go with Grokking the Coding Interview to learn some patterns that gives you good coverage for questions https://www.educative.io/courses/grokking-the-coding-interview. If you want to go with a bootcamp, I recommend Interview Kickstart, but that's like a 2-8 month commitment.
|
# ? Jul 26, 2021 20:53 |
|
Grokking is definitely good, but I'd for certain go with https://www.educative.io/courses/algorithms-ds-interview. first and play it by ear with how much more prep you want to do by touching on harder problems in each area outlined there plus Grokking. The amount of breadth that covers in terms of what you might actually see is a lot better than Grokking, which digs super deep in certain areas.
|
# ? Jul 26, 2021 21:36 |
|
I was really excited when a recruiter offered to poach me from my current contract for a 33% pay hike. But then I read the fine print of the contract and one of the clauses allows them to fine me for "three (3) months of the compensation paid" if I don't report every solicitation for a new job within 2 days. Not only did they just try to poach me, I get 5-6 (mostly spam) emails about new jobs a day. It's only a 1 year contract with at will termination by either side, so that's a quarter of my income in the best case. Besides which, it's phrased in such a way that I can't start job hunting for my next contract until after my current one ends without risking losing the last three months pay. I think I'm going to call the recruiter back tomorrow and see if he can reduce the penalty clause to cause for termination. If not, I think I'll pass. I understand wanting to prevent contractors from converting to direct hire in the middle of a contract, but three month's pay is absurd and the scope of actions that can trigger the penalty is far too broad.
|
# ? Jul 26, 2021 23:41 |
|
I would never work for a company that tried to pull that in a contract, because they're going to try and screw you over in other ways just as badly.
|
# ? Jul 26, 2021 23:49 |
|
LLSix posted:I was really excited when a recruiter offered to poach me from my current contract for a 33% pay hike. But then I read the fine print of the contract and one of the clauses allows them to fine me for "three (3) months of the compensation paid" if I don't report every solicitation for a new job within 2 days. Not only did they just try to poach me, I get 5-6 (mostly spam) emails about new jobs a day. Malicious compliance version - actively report any and all positions ever seen. Write an app to quickly report every single recruiting ad that hits your eyeballs while browsing the internet. Only do this during work hours and also actively search and report for positions that pay more and are enticing you by existing. Even in your imagination. Watching television and a recruiting ad comes up for local warehouse? REPORTED! Someone at the bar mentions an open position? REPORTED! Actually just forward that to me, I'll do that job better than they've ever seen.
|
# ? Jul 27, 2021 04:02 |
|
What's the best way to "start" with Amazon? I went ahead and used the hired.com lead. Had a bunch of linkedin messages too. Feel like they have too many postings for me to pick one and apply myself. I know the usual caveats on them having a really bad work-life balance I'm just keeping my search broad.
|
# ? Jul 27, 2021 14:08 |
|
Pollyanna posted:Lord help me, Google's interested, and now I'm staring at a pile of prep materials. You know, I would even let you skip the takehome if you applied to my listing
|
# ? Jul 27, 2021 19:47 |
|
one day i will remember everything i need to do but that day is not today
|
# ? Jul 27, 2021 21:32 |
|
Heard more about the unfunded start-up over lunch today. Seems they're close to getting funding from a big and experienced investor. I heard their business plan and I think they can easily sell the product once it's built. Really like that it's a b2b product that would generate real, quantifiable value rather than being a consumer app that hopes to win the adoption lottery. Pretty straightforward automation that would save their target customer a lot of man-hours and pay for itself by letting them do more work without more employees. From what I've read about marketing over the past couple years, their product would be an offer that their target customer would be stupid to turn down, since the value is obvious and the price is quite low by comparison. Seems like the sales plan relies on the founder's personal connections but the founder is extremely well connected in their industry, has deep experience in it, and wants to build a product to solve a specific problem that is 1) currently unaddressed and 2) they currently receive calls for advice on how to handle. Honestly surprised that my friend knew a guy who knew this guy. They want me to build and lead a team to build the product. Seems like a good opportunity and I could see it comparing to some other large b2b companies that went over $1b in valuation but it sounds like a 60+ hr/wk job for the next five years. So it's 60+ hrs/wk for uncertain gain but probably the best odds a startup could have vs a chill 35hr/wk job at Google where I will almost certainly become a millionaire again in five years just by relaxing and enjoying my free time. If I wasn't already tired all the time despite being unemployed and didn't expect a stock market crash within a couple years followed by a couple decades of economic depression, I'd be really excited. I just don't see new unicorns being minted in the next 5-10 years. Will see if they come back with an offer and how much equity is in it. Maybe that'll change my excitement. Right now I am tempted to just take the easy job at Google and start a side business if I later feel like working so much. oliveoil fucked around with this message at 01:28 on Jul 28, 2021 |
# ? Jul 28, 2021 01:25 |
|
i really don't think you should do any more gambling
|
# ? Jul 28, 2021 05:47 |
|
It sounds like you want to be talked into taking this job. Doesn't sound fun to me but I'm pretty risk-averse so Basically Achmed Jones posted:i really don't think you should do any more gambling
|
# ? Jul 28, 2021 07:24 |
|
the way you talk about the founders makes you sound unfortunately naive do not trust founders promises and do not trust vc promises and dont get excited even if it seems exciting, hth and dont go into startupland if you have a gambling problem. its a big ol game of positive-EV poker
|
# ? Jul 28, 2021 08:00 |
|
oliveoil posted:Heard more about the unfunded start-up over lunch today. I mean, if you're bored, go for it. Sounds like a W2 consultancy job to me. Tons of unicorns, I think my last company merged with another startup and now they're a unicorn. My new company went unicorn in 2019 Doing chillax stuff and becoming a millionaire sounds like more fun, personally. I'd rather work on growing my vineyard or whatever, rather than work myself to the bone at a startup for peanuts with no economic moat
|
# ? Jul 28, 2021 08:06 |
|
Yeah, when you guys put it that way it sounds less appealing. Wonder if there's a way to back out without pissing anyone off. They asked if I was interested in being considered and I said yes, but I suspect I'm the only one being considered as they said they can't get funding until they find a tech person to join. Going to feel bad if they go tell people they've got an engineer, get funding and then I turn it down.
|
# ? Jul 28, 2021 12:16 |
Why you would ever consider a 60+hr/week job is beyond me to be honest.
|
|
# ? Jul 28, 2021 12:21 |
|
I went from 37.5 to 40 recently and even that I'm regretting.
|
# ? Jul 28, 2021 12:22 |
|
oliveoil posted:Wonder if there's a way to back out without pissing anyone off. gently caress em
|
# ? Jul 28, 2021 12:29 |
|
There's nothing appealing about that imo - we totally are going to get funding (maybe) - you totally will work 60h weeks for the next 5 years (but we might reward you for it then! maybe) - our product is super viable and there's definitely a huge market for it (said literally every startup CEO, ever)
|
# ? Jul 28, 2021 12:31 |
|
As everybody here is saying: sustained 60hrs/week is the burnout territory. If they are asking you for that kind of commitment from the start, I'd really questions the "market fit" they are thing to sell you.
|
# ? Jul 28, 2021 12:40 |
|
This sounds like a bad idea.
|
# ? Jul 28, 2021 13:19 |
|
Yeah that’s not a unicorn, that’s your bog standard everyone-thinks-their-idea-is-flawless-so-you-should-work-long-hours-for-free ugly rear end horse. Nothing remotely special about it.
|
# ? Jul 28, 2021 13:20 |
|
|
# ? May 23, 2024 20:23 |
|
gbut posted:As everybody here is saying: sustained 60hrs/week is the burnout territory. If they are asking you for that kind of commitment from the start, I'd really questions the "market fit" they are thing to sell you. They're not asking for 60hrs but I feel like it would probably be necessary in order to stay ahead of the competition since the thing is obvious once you see the need for it and could probably be built in a few months if you already had a team. If you've ever worked a non-tech job and realized that a long, annoying manual process you do all the time consists of a bunch of smaller tasks that can all be easily automated, but nobody has done it because nobody doing it is an engineer, that's basically what is going on, and apparently it's a thing that lots of people at lots of places are doing. I.e., there's no moat that I can see as another said. I'm not sure if the founders noticed this (probably not since they're non-technical) but speed is likely the most important thing here. Once companies already serving this industry notice that this need is unaddressed, it'll probably be easy for them to add it. I think they'd need to constantly add to their product to solve new problems in order to stay ahead. rt4 posted:gently caress em I'd agree with that usually but this is a good friend of a good friend and they're planning to quit their jobs as soon as they get funding, which is only blocked right now by finding a tech lead that they trust. I guess I'm just worried they'll come back with "hey we got funding and quit our jobs, let's start!" oliveoil fucked around with this message at 13:54 on Jul 28, 2021 |
# ? Jul 28, 2021 13:47 |