thegooley

Microsoft Research, Immunomics | Seattle, WA | Full-Time, Onsite | https://aka.ms/immunomics

Position: Senior Software Engineer (Data)

Overview: The Immunomics team at Microsoft Research is working to decode the human immune system to better understand and diagnose disease. Our effort is called The Antigen Map Project and we're partnering with computational biologists and immunologists to combine best-in-breed biotech and machine learning techniques to tackle big problems.

We're a small team operating inside Microsoft Research as part of a larger healthcare innovation effort and we're looking for another software engineer who is excited by our mission, loves building software to make the rest of the team more productive, and is comfortable operating in a startup-like style.

What you'll do: One year into the project and our data volume is increasing rapidly. We're looking for someone to create and own the data pipelines. Beyond working with data, you will have opportunities to contribute to platform libraries, work on full-stack applications, and more. Bio/ML background not required - but interest in learning is a must.

Our tech stack at this point is primarily Spark, Scala, and Python running on Azure.

Apply directly via Engineering link at https://aka.ms/immunomics and/or contact gooley@microsoft.com if you have questions!

Damn! I should have patented it!

Yes, please steal it. We debated whether it would be a good idea to do API-only registration, but the response so far has been mostly really positive.

We actually do have a JS API [1] but we wanted to focus on a quick demo for developers rather than an integration wizard. I think it's important for people to understand the premise and how it's a bit different than other "analytics" products and just logging pageviews ala Google Analytics will give a sub-optimal experience.

We need to put a lot more work into the onboarding process (we created the whole public site and api docs and demo wizard over the weekend) so if you have any good examples of "getting started" flows that make sense, please send them my way.

[1] http://lessneglect.com/api (and select the Javascript tab on the code samples)

We struggled to really explain the value prop in a 2 minute video, for sure.

The concept of proactive (as it exists in our beta) is more in the sense of "fix their problem before they have to ask you to fix it". But our goal is absolutely to help you to "fix their problem before they know they have a problem".

We've been able to do that to a certain extent with FolioHD customers manually by seeing patterns in their usage where we can suggest an optimization.

For instance:

We see a user uploading images one by one and can shoot them a message saying: "hey john, did you know that you can select all your images at once when you're uploading? save some time!"

or, we see someone uploading custom images to use for their homepage slideshow that are branded with logos that don't look great in their portfolio proper, so we can reach out and say "hey jamie - did you know you can create a separate gallery for your homepage features and then hide it so it doesn't show up inside your portfolio?"

So this is what we mean when we say "proactive". When we see one or two people having problems, we can help them out to best use our product. And if we see a lot of people missing key features, we know that we need some UX work.

It's been pretty great, and that's why we figured we should polish it up and release it for other people to use.

And we look forward to accepting your money!

Thanks for the feedback, and if you have any specific ideas of how we can better explain it - please shoot them my way. And yeah, we'll see about getting rid of the marketing-speak in our next update.

Sounds like you've groked the big concept, but we'd love it if people didn't have to ponder over it before understanding what it is. Thanks for taking the time.

Awesome! Can't wait to see it, and let me know if I can help out on the API side at all.

Sounds plausible to me. I suppose it doesn't matter how fast the SSD is... it has to restore (up to) 16GB from disk to memory before it's ready.

I haven't seen this particular issue, but I have had a whole bunch of strange, likely firmware-related issues on my Retina.

For instance, sometimes the screen just won't shut off when trying to put it to sleep - Closing the lid and the screen/logo stay lit up indefinitely. Chrome and the integrated graphics card really don't play well together - it's caused an OS reboot 10's of times in the past month. Mostly, waking the computer from sleep takes a fraction of a second but sometimes it takes ~20 seconds to show the screen.

I would guess that all of these are related to the integrated Intel graphics chip firmware/drivers being not quite baked.

BUT - in spite of all of that, this is without question the best laptop I've ever owned. It's also the first laptop where I haven't wanted or needed to use my iMac anymore. It really is finally a true "desktop replacement" that is totally portable. And I could never go back.

Looks like it was heavily inspired by this (from ~2010): http://culturedcode.com/status/

And I really like the concept, but I believe that the value of a "status dashboard" is that it's a concise view of what's happening right now. I really don't think that this sort of concept is a good resume replacement it's not a good format for 20+ items.

I'd suggest making it a status dashboard for what you're working on right now, and once things are "completed" they drop off into a bulleted list at the bottom which becomes your "work log" rather than "dashboard" at that point. Make it less overwhelming and I could really see a recruiter or business partner getting some good info from it.

Ah yes, this method does set the height of the header and footer as constant.

Is this the kind of code that you're looking for? http://files.thegooley.com/example-fullpage-app.html

Nice phrase, I'll have to keep that in mind - at least when talking to hackers.

Re: replacement of Facebook, I don't think it's realistic for someone to develop an actually decent federated replacement for a true social network. The difficulty of doing some of the stuff that Facebook users like, but trying to make it work when all of the nodes are on different networks with different constraints, seems to be very high.

I'm not so much in the "Facebook shouldn't have my data" camp... more like the "I want control of my all data TOO" camp.

Our concept for people writing apps that leverage TheMux doesn't require them to actually load data live from their user's instance of TheMux (because that's crazy, a single point of failure, and probably very inefficient) but that there is simply the agreement to interchange data.

To steal your phrase - we want to make the data symmetric.

So that no matter where I create or update my data, the systems involved makes a best-effort to return that data to my MuxDB, and then my MuxDB will make a best-effort to get that data out to the sites I've authorized and who need the data to do what I've signed up for.

Obviously, we're still in the early stages, but it's been an exciting and challenging journey so far.

It's interesting that people seem to have suddenly started talking about data ownership much more lately.

We have been working on a project called TheMux [1] that aims to create a data platform (for lack of a better term) which helps you to pull in and archive your data from various sources and apply some simple normalization. We're specifically working on ways to keep both content (blog posts, photos, status updates, etc) and datum (health info, workouts, communication data, etc) in a form where you a) have full control of the raw data and b) can make select data available to external apps which do things like presentation and analysis.

Our goal is to create an open-source platform to address some of these questions around data ownership, access and portability. Imagine the day when you decide that you want to move your blog from Tumblebook to Posterpress and you can do that by simply creating a new account on Posterpress and granting it access to your existing data. Or you've been using JogKeeper but then you find this great new service called SprintTracker that you want to try out - and all you need to do is connect and it will have your years worth of running data.

And we think that something like this will also make it much simpler for SaaS developers to compete not by customer lock-in but rather by providing superior products and continually working to make the customer happy.

We're taking it slowly right now to build this platform which we will open-source under a permissive license (as soon as it's a little more mature) by first building a few consumer-oriented services on top of it. Number 1 on our list is a blog-type website based on this MuxDB concept and that's what we're working on at the moment.

If you're interested in giving us your thoughts (or help, or tell us we're crazy, or whatever), my email is in my profile.

[1] http://theMux.com

I've been using Backbone at Earbits (YC w11) [1] since February and have been very happy with it. We're running a pretty complex AJAX app that streams music along with providing a large set of functionality around the music itself (artist info, local concert alerts, track history, deep facebook integration, bookmarking favorites, event logging, running contests and promotions and so on).

For an idea of scale, we have: ~60 backbone views, ~120 javascript templates (underscore jst), ~30 backbone models + collections, 1 controller (the router)

And we have a fairly complex system of white-labeling that allows us to swap out javascript templates and augment view functionality for partner sites [2] and local sites [3] without adding a lot of overhead.

The combination of Backbone + Underscore + Jammit with Rails/Rack on the backend is working great. Even developers who are not familiar at all with Backbone and the ajax app, can dive in and understand everything pretty quickly.

Can't possibly recommend it highly enough, if you 1) are comfortable with Javascript, and 2) prefer to roll most of your own code so that you're not beholden to bulky rigid frameworks.

There are some quirks that annoy me, but overall I can't complain. One of these days when I have some more free time, I'll try and write a series of posts explaining how we're using it and our key learnings.

[1] http://earbits.com [2] http://sfgate.earbits.com [3] http://la.earbits.com

Sure, but I bet you could still look at all 10 photoshop manipulations and say "yep, those are all based on that original photo".

An interesting viewpoint, but I think the differentiation here lies in the degree of "artistic" creation. Put 10 photographers in a room with Miles Davis and you'll get 10 pretty distinct photographs of the man. Have 10 people create a pixelated version of a photograph, and you'll have 10 pretty similar results.

Creating a pixelated version of this photograph is more akin to taking a photograph of the Mona Lisa - sure you can get a cool angle or do it with infrared film to get a different color palate, but it's still a picture of someone else's original work. You wouldn't say that the Miles Davis cover photo is "based on a Miles Davis concert", but rather it was created _at_ a Miles Davis concert.

Similarly, in music, remixing a song can be a very creative process - but it is still a derivative work in that it wouldn't exist without the original which was created "from scratch".

(Disclosure: I am/was a photographer who used to earn a living solely from licensing my photos for editorial purposes)

Edit: I'm not attempting to discuss copyright laws (which are pretty clear), but rather the philosophy and justification behind the creation and copying of artistic works.

I understand the hesitation, we try to avoid asking for anything that we don't need to make the user experience better. We don't post anything without you explicitly performing a "share to facebook" action, so you can rest easy on that front. And we ask for your profile info so that we can see what bands you already like.