Hacker News Reader

Comment by stephensonsco | Hacker News Reader

These are definitely great points. Most companies looking for DL/ML talent aren't interested in setting up HR hoops for the applicant to jump through.

They want to see if you did cool stuff before you applied for the job. If you didn't then you won't get an interview, but if you did then you have a chance no matter what your background is. Of course, the question of "what is cool stuff?" comes up. If it is building small projects with a a little bit of success, that probably won't do it (it might work for larger companies, or companies that need light ML/DL performed). But if it is "built twitter analysis DNN from scratch using Theano and that can predict the number of retweets a tweet will get: here's te accuracy, here's a link to a my write up on it and here's a link to github for the code.".

Edit: added words similar to this at the end of the blog post.

View on HN

> So even if you're a beginner with deep learning, you're welcome to apply for one of our open positions

Statements like this contradict what you are saying here - to really build a model that predicts number of retweets based on the content of the message (not something like the average number of retweets this user has) is very non-trivial. If your threshold of a side project is publishable [1], it is an unrealistic expectation.

[1] http://homepages.inf.ed.ac.uk/miles/papers/icwsm11.pdf

Speaking for the other side, the point is not that you achieved the accuracy but it is understand how you thought about the problem. The ability to decide on a useful hypothesis, formulate the problem around it, and have some way to measure your progress is very very valuable from the employer's perspective.

Sure, my point was more that this can be demonstrated by a much simpler project. How many candidates have you seen who haven't had a deep learning job before complete such a complicated project?

If you:

- have good problem solving and coding skills - spend a month or two learning how to build networks using good libraries

Then you will be able to get a good result with the Twitter task I pointed out. It takes being able to work input data correctly, think about what matters and what doesn't, then synthesize using readymade DL tools, usually in python. None of that is complicated esoteric neural net stuff, it is just motivated problem solving.

I'll add that as a big point — you should be able to code, problem solve, and be motivated.

The important part to mention is that the expected completion time of your project is 1-2 months. I have extensive ML experience (no deep learning though), and I think this project would take me at least 2 months to do well. This is a long time, especially for a beginner side-project. It is good to tell people to do side-projects, but suggesting projects like this is de-motivating for beginners. If someone on your team (who is presumably an expert already) does this side-project I'd be very interested in the results, and on how long it took them - it would make a good blog post. Your proposed beginner side-project should be less complicated than a published paper with 200 citations (the paper didn't use a nn).

Everyone thinks they can download tensorflow, hook it up to some stock market or Twitter api, hack a system in a week, and become a ML engineer. It is simply not this simple. It takes many months or years to achieve this experience.

Sometimes software engineers need to accept they aren't the smartest ones anymore. This is why they can't get the "smart" and "cool cutting edge" ML/DL jobs.

Reply on news.ycombinator.com