My thesis work focused on scaling recommendations to millions of users. Through that work I got a different perspective on personalisation. For better or worse I’m not yet sure. Tonight I received this e-mail from LastMinute.com which got my attention.
On the other hand the touted master of recommendations, Amazon, didn’t get it quite right in their last e-mail. Their recommending me a book I bought just two weeks ago.
LastMinute, you did it right this time.
I wouldn’t consider myself much of an Apple-fanboy, but they do usually put on a good show and it’s fun to follow the discussions afterward. The last two times it has been about phones, it hasn’t, unfortunately, been living up to my expectations.
As I read the live-blog transcript I couldn’t but help noticing the lack of anything new (confirmed by this Mashable article). Is Apple playing catch up with their competitors for the first time in their mobile-history?
One particular technology improvement I would have liked to see is the ability for the mobile to communicate with local devices at a short range. NFC is one such technology. Whilst several of Apple’s competitors already feature it in their phones, the market overall doesn’t seem as enthusiastic about the technology or its potential. An Apple integration would have conveyed a strong support for the technology possibly sufficient to spur a plethora of new services and businesses.
The fact that they aren’t supporting might suggest they have other ideas (better ones?) in mind, or maybe we’ve grown too used to small miracles when Apple presents?
Been busy relocating to Malmö lately and haven’t really got the time to sit down and tinker with something in a while. Tonight that changed as I had stumbled upon human.io earlier today. Human.io is a simple way for developers to write and deploy small mobile applications, such as surveys.
I’ve been toying with the idea of reducing the need to install apps on your smartphone for quite a while, but haven’t seen any great examples lately. Apps, in my opinion, are a pain in the —-. Ideally the information “should just be there when you need it” (TM). That is easier to say than do though.
Human.io simplifies alot of things. Through an intuitive python API you can hack together a mini-survey in no time. Using for example geo-location information you are able to restrict the survey to people in a specific area. In my case, people in Västra Hamnen. There are however a number of things that can be improved:
Its great to see movement in this area. I’m looking forward to see what more these guys come up with as human.io progresses.
Tasty!
Every once in a while you reach a point where you know a lot of things are about to change drastically. I’m on the verge of such a point. Which, admittedly, is super-interesting and exciting!
About a month ago I successfully defended my thesis called Online recommendations at web-scale using matrix factorisation, the last stop of a two-year journey called European Master in Distributed Computing. Supported by the Erasmus Mundus scholarship, this little endeavour took me to Portugal for one year of intense course work at Instituto Superior Tecnico. Together with four supersmart classmates from India, Pakistan and Brazil we tackled courses on parallel and distributed computing, virtualisation, mobile computing, cloud computing, network security, P2P and overlay networks, and fault-tolerance. We literally dived into a world of wonderful complexities and got to work on some really interesting and challenging problems through various course projects (a bunch of them can be found on my Github).
The journey continued the second year at KTH in Stockholm for another semester, diving deeper than before in the waters of distributed computing and systems. Joining up with our classmates from Barcelona, an already sparkling enthusiasm became raving ecstasy! Prominent guest lecturers from EPFL (Lausanne, Switzerland), UC Irvine (USA), and NTU (Singapore) gave talks on distributed databases, mining mega-graphs, and coding theory. These talks were complemented by industrial talks from Google, Spotify, Klarna, and Nasdaq. To name a few.
During the fourth and last semester I moved to Barcelona and the (nowadays international) social network Tuenti for the final piece of the puzzle. After a code test and three pretty hefty interviews, they signed me on for working on recommendation engines. Another deep dive in machine learning techniques and six months of pure awesomeness followed. Seeing what my colleagues at Tuenti achieved during this time gave a refreshing insight on how fast-paced and fun (and naturally daunting) the industry can be from the first row.
After a month of backpacking in Chile, Bolivia and Peru, I’m returning home to Sweden as “one happy monkey.” Surely there are things which were not perfect during the programme; particular courses can be improved and administration can at times be smoother, but the experience has been invaluable. Anyone looking for a master’s degree and is ready to put in just that little extra effort needed for an Erasmus Mundus experience won’t regret it. I promise.
In a few weeks I’m moving to Malmö and I’ve intensified my search for new adventures. Lets see what’s around the next corner!
Arrested in June 2011, Mr Cleary was released on bail but later jailed after breaching his conditions.
He had been banned from accessing the internet, but is said to have contacted hacker Hector Xavier Monsegur, known as “Sabu”, who had been working as an FBI informant following his own arrest.
Writes BBC with regards to the Lulzsec hackers that were arrested for having attacked several government and private (power) institutions.
I wonder, how do you ban someone from accessing the Internet these days when it almost is omni-present? Send them to the deep forrests of Bolivia or a remote desert in northern Chile? That’s two places where I’m going. I’ll be back in a month.
See you!
Diego, a colleague of mine at Tuenti, is coordinating short 5-minute presentations that we conduct here in the office at Tuenti every second week. I love the idea of being able to prepare short, funny, interesting presentations and listen and learn from what others care about. Naturally I wanted to contribute too, and hence went with something that I feel all-to-often is under-valued in our field (and probably in many others too), presentation design.
Having done a similar presentation before I opted to do one again (and because I like presentations and want to push Diego’s cause) and copied a lot of the material into this one. In the end however the visuals and content ended up quite different from the first one. For one incredibly important reason: the audience.
Last time I did the presentation on presentations (very meta) it was as part of a university course in scientific writing and communication. Part of its curriculum addressed presentation skills, including structure, language and tone, delivery, and a tiny bit on visuals. In other words, during my presentation I could assume that my classmates already grasped the basics. In the case of my colleagues, I could not make the same assumption and therefore the presentation in the end became quite different.
In five minutes you cannot, and should not, cover too much. Perhaps I did try to squeeze a bit too much content in there and it got a bit rushed. As I try to make a point not to run overtime (so should you) there is a lot left to be said. Overall, however, I’m pretty happy with it.
Next week I’m doing a longer tech-talk where I present my thesis work.
Since I started writing my thesis on building a scalable recommender system at Tuenti I’ve been using Zotero to manage papers. And so far it is awesome!
Most digital publications are recognised directly by the Zotero-chrome plugin and so there is never a need to manually import a reference. If I download a paper I immediately link it with the reference in Zotero so that I do not have to remember which paper belongs to which reference. For some reason the digital archives have weird naming conventions so this makes finding the right paper again a breeze.
Of course Zotero has support for exporting the reference list as BibTeX, hence, I’m all set for a much more pleasant writing experience compared to the overly complicated office suites (OpenOffice.org and MS Office).
Thesis writing toolbox as of today:
You should try it too.
Mimmi, a friend of mine, recently launched her own company under the title “En promenad om dagen” (freely translated to “A walk a day”). And her idea is simple: exercise thirty minutes a day and you will be rewarded.
The second thing I admire with Mimmi’s initiative is the fact that she is doing it. A lot of people talk (including myself many times) about doing something about something, but never or extremely seldom do it! I’m looking forward to see what experiences she will learn and how it will progress. In the mean time, stay up to date by following her blog.
Investors, venture firms and start-up coaches always preach one and the same thing: keep it simple. It’s their mantra. And yet keeping things simple is probably the most advanced art form around. I’m fascinated by it. Most things are rarely simple at first. Only over time, and by iterating and evaluating an idea, a report, a software program, or business plan, are we able to discover its simplicity. I believe Mimmi is on to something.
Det bästa företaget är det som blir av ;)
Keep going!
Next week I have an exam in Spanish. While I think I am making progress I still have a hard time remembering words. Haven’t used word lists since 4th grade I guess. Today I have a new take at it.
Essentially it is down to practice, so I wrote myself a little application to shoot words at me and correct me if I’m wrong. I’ve built the list on my own based on words from our classes but it can easily be replaced with any other list.
Maybe tomorrow I’ll convert it to a web-app so that my colleague can rehearse too.
Repetition is the mother of all knowledge.
Update: It now keeps track of my score too, quite discouraging at the moment. Looking forward to changing this!
I recently had the chance to review a project audit of great length. One of the key contributions (if I may call it that) in this report was the importance of aiming for simplicity. Something which certainly isn’t simple.
For me, building new software can be extremely cumbersome and all too often I find myself implementing things features which are not needed yet and, hence, dodging simplicity. To be completely honest, sometimes I don’t even know if they will ever be needed, and still I spend time hacking them out. A beginner’s mistake perhaps, yet utterly difficult to realise when you’re in code-monkey mode.
When we developed the website for the 22nd World Scout Jamboree 2011 we started out with the basics, and over two months developed the majority of things visible on the site today. It is not maintained anymore as the Jamboree is over. During its life, however, new features were added as a response to the buzz we recieved on-line. One example is the Social, or Live, page which would display what Scouts were saying about and doing at the Jamboree right now. This feature had two incarnations: first as Social which simply displayed Facebook, Twitter and Youtube feeds. Later, as Live, which integrated a live-blogging service used at big events, a twitter feed of #jamboree2011, and our Bambuser channel.
My very first sketch of the Social tab was extremely simple and only aimed to convey the idea. It looked like this:
Some say failure is the only way to progress. Lukas and Annie who reviewed my wonderful drawing didn’t really understand what all the boxes were and so after its first review, and me frantically trying to master Gimp, it looked like this:
Still boxes, but it was encapsulated in what we already had in place. Eventually, the Live tab came to this:
What started as a simple idea and a few silly sketches was with the help of an amazing team transformed into one of the most viewed pages of the Jamboree website.
Starting simple (and boring and ugly) but enough to convey the idea is probably the most underestimated concept ever.
Now, what do you do to avoid the conquer-the-world-all-at-once-trap?
