Reddit is hiring engineers in ads ML for our brand-new NYC office. We’re looking for both junior and senior ML engineers.

The ads ML team is responsible for all the machine learning that powers our ads backend, as well as the inference and ML serving infrastructure.

Our backend is in flux, but we’re using: Spark and Scala on AWS for machine learning, as well as BigQuery on GCP for analysis.

Our ads serving stack is written in go and runs on AWS, and our ML includes logistic regression, GBDT, NLP embeddings, recommender systems, and more.

If you’re interested, shoot us an email at jobs+hn@reddit.com

ML engineers fall between data science and production engineering. They're responsible for doing their own data analysis and experimentation, plus some light engineering to implement their models in production. We also hire backend engineers to handle the more hardcore infrastructure tasks, so we don't expect quite the same insane level of coding ability as some companies do.

Thank You

Reddit is hiring engineers in ads quality for our brand-new NYC office! We’re looking for machine learning and backend engineers, both junior and senior.

The ads quality team is responsible for all the machine learning that powers our ads backend, as well as the inference and ML serving infrastructure.

Our backend is in flux, but we’re using: Spark and Scala on AWS for machine learning, as well as BigQuery on GCP for analysis. Our ads serving stack is written in go and runs on AWS.

If you’re interested shoot us an email at jobs+hn@reddit.com

Sometimes there is an explanation (e.g. the flight doesn't exist, or has a 30hr layover), and other times we don't have the flight, in which case we want to alert our providers to fix it.

- Consolidators [that you don't have commercial relationships with] have wholesale arrangements with airlines allowing them to sell tickets below the direct/GDS price. I suspect you're more reliant on ITA results than many other portals or meta-search sites and their prices are trivially easy to beat on competitive routes if you have a broader range of partnerships.

- Some consolidators put together logical itineraries not bookable via GDS or even direct with the airline. As an extreme example I once bought a cheap, logical pair of connecting long haul flights with Saudia via eBookers that Saudia's website wouldn't allow me to book at any price direct with them; I'd guess the airline didn't want to market anything involving a 5 hour layover in an under-redevelopment terminal in Riyadh. There are plenty of less glaringly obvious potential indirect itineraries that consolidators (potentially including Kayak as well as their affiliates) can identify with Innovata's minimum connect time file and schedule data[1] and other providers might have a more liberal view on reasonable connections than ITA

- Sometimes other sites' metasearch results' advantages are purely illusory, based on outdated results they've cached or screenscraped. I find Skyscanner LCC prices are frequently out of date, even though the actual ticket price is also usually cheaper than ITA's. Other sites might also create the appearance of more choice by not combining codeshared flights into one.

As for LON-SYD on Monday Kayak has a lot more results under $1000 than you. It's possible their much cheaper ticket prices with SIA are outdated or geolocation specific, but the cheapest flight they show (a Garuda two-stopper that's no more obviously agonising than your cheapest result but 10% cheaper) doesn't show on your engine at all. Then again, the price with Kayak had risen to only 5% cheaper than your best price when I started the booking process...

[1]disclosure: I sell that data, amongst many other things, although not to OTAs

But when I clicked on the button to book, the price was different (higher) on the airline website.

There are quite a few comments here complaining about the gaudy unusable interface produced by "modern JavaScript-based techniques", which in this case appear to provide virtually nothing that static HTML and CSS can't do. The dev team was lazy and wanted to show off, and screw anyone who doesn't think running gobs of arbitrary code should be a necessity for buying a ticket.

And for what it's worth my non-developer artist girlfriend has been using noscript for longer than I have. I'll let her know what modern devs think of her, though I don't think she's heard of wget.

I actually deliberately didn't use Django specifically because it's so complicated for beginners (and for pros, IMHO). I was trying to avoid magic. Webapp2, which is the web framework GAE users, and what we used in the course, is about as simple as you can get without starting from scratch.

That said, the webapp2 decision is by far my biggest hesitation in recommending it. Webapp2 just doesn't have the community support; whether django is too "heavy" or flask too "light", they're both popular, which means they have great support, plugins, help on SO, and a lot of developers familiar with them.

Webapp2 also hasn't seen a release since 2012[0].

I write this having recently completed a small consulting project where I helped a team on GAE move from Webapp2 to django. The productivity gains just from using django plugins and builtins was insane. They had been on webapp2 because their dev learned webdev on your course, and I think he would have been much happier had the course included a more widely used framework. Lock-in on Appengine is also a major issue; the work to move off can be enormous, especially if you're using a framework designed for GAE.

Again, I totally understand the decision to use webapp2 and the course is still a huge net positive. But if you ever think about revising it, I'd recommend considering another framework, and possibly even using Elastic Beanstalk.

EDIT: I understand django can take a little more to get up and running, but most people learning Python are going to have to learn it very soon anyway. I learned how to program by doing the django tutorial, and didn't find it burdensome at all.

[0] https://pypi.python.org/pypi/webapp2

I personally don't like working in frameworks of that size because I find myself fighting them to do what I want.

For the course itself, I really wanted the students to get as close to the server as possible. Then, when some of them switch to the heavier frameworks, they'll have a better idea of what the framework is doing and what is possible. The course had a time constraint on it as well.

All in all, your anecdote reenforces my feelings on the topic :)

Whatever people might say about webapp2 and App Engine, I think they were the right choice for the course. I had no problem moving on to other frameworks or moving away from App Engine, having gotten a solid understanding of the concepts from your course.

Once again, you have my sincere thanks.

The first thing we added was the "rising" page, which used to be reddit's default "new" page. The rising page was a weighted new page. It was a little hacky, but worked for a while. I see they've changed this, which I think may have been a mistake, but I haven't thought about it in forever.

The second thing we did, which worked really well, was to have an up-and-coming link placed at the top of the list on the frontpage. This helped those borderline posts get more visibility. I had a reddit simulator I used to use to test things like this. That space appears to be a subreddit search box for me, so it seems they've moved onto another solution.

Randomness is an interesting idea, though it might be a bit difficult to fit into the way we cached things.

I think the problem that the randomized algorithm solve is not missing good content. It is to give the same chance for equally good content - or at least to minimize the error due to page limit.

to have an up-and-coming link placed at the top of the list on the frontpage

Two things: 1) either the up-and-coming are a few borderline posts taken in order and then here you'd just extended the front page with a few more posts (ie. it changes nothing) 2) either you'd carefully choosen a heuristic for those posts which is pretty much what is suggested here.

Anyway, thank you for sharing your experience with this problem, it's very interesting.

An other idea would be to fix the content quality and make variable the number of posts in the front page.

For example: http://www.hipmunk.com/flights/BCN-to-CDG#!dates=Dec31

Good stuff. Saves me the time from matrix.itasoftware.com to go to the airline or booking site.

What about limiting it to fare basis? for example: sfo::/f bc=y atl::/f bc=y doesn't work. There another way?

Love what you guys do and will have to check on working there at some point.

I know you guys can win this, best of luck! :)

It would indeed be ridiculous if someone had a patent on displaying data in multiple computer windows, but that's not what was patented.

Here's claim #1 -- the other claims are similar in their specificity:

1. A system for synchronizing the presentation of data on a digital computer display, comprising:

first and second window-controlling means, each of said window-controlling means displaying information in at least one display window, a synchronization control means, and means for communications between said synchronization control means and each of said first and second window-controlling means; wherein said first window-controlling means displays first information from a first source, and said second window-controlling means displays second information from a second source; wherein said first information and said second information have sections, and at least one section of said second information corresponds to a section of said first information; wherein said first window-controlling means sends a message to said synchronization control means over said means for communications indicating a change in viewing position to a new section of said first information; wherein said synchronization control means sends a message to said second window-controlling means over said means for communications requesting a change in viewing position to a section of said second information corresponding to said new section of said first information; and wherein said second window-controlling means displays said corresponding section of said second information on at least one display window.

Remember, patents cover methods, not purposes. This patent covers a specific method for displaying data in multiple windows, not the general concept of displaying data in multiple windows.

Mind you, I still think it's a dumb patent. But let's make sure we understand what we're lynching them for before we get out the torches and pitchforks, ok?

- Two or more windows displaying data from different sources

- The sources have sections

- Sections from one source have corresponding sections in another source.

- Changing sections in one window makes other windows change to corresponding sections.

- Basic message-passing communication to implement the above.

[1] http://www.ics.uci.edu/~redmiles/ics227-SQ04/papers/KrasnerP...

I wrote a split screen sprite editor for dos about 20 years ago. I bet it's still on a 720 floppy some place in my basement. In fact, I think I have a source printout filed away too..

Have you ever used Hipmunk? It's user friendly and all, but it's not rocket science in the 50s. I think it's fair to assume that any patent they're violating is probably bullshit without too much research.

    method and system for synchronizing the presentation of 
    data from different, but related, sources in different 
    windows of a computer display