LinkedInAapo Kyrölä

Aapo Kyrölä

Software Engineer at Facebook

Pittsburgh, Pennsylvania
Computer Software
  1. ProtoGeo,
  2. Carnegie Mellon University,
  3. Twitter
Recommendations1 person has recommended Aapo

Join LinkedIn & access Aapo's full profile

Join LinkedIn & access Aapo's full profile. It's free!

As a LinkedIn member, you'll join 300 million other professionals who are sharing connections, ideas, and opportunities.

  • See who you know in common
  • Get introduced
  • Contact Aapo directly
View full profile



Startup entrepreneur, computer scientist and engineer with 20 years of experience. Strong background in leading and hiring software development teams at two fast growing startups; building scalable distributed systems; algorithms and data structures; machine learning and related plumbing. Recently I have obtained valuable experience in real world machine learning problems (instead of those clean academic problems) during my visit to Twitter and my work at the activity recognition technology of Moves app.

Founder of two pioneering startups: Habbo Hotel (2000, web-based virtual world, reached 75m$ annual revenue) and Moves app (2012, fully automatic activity and location tracker for iOS). At CMU, for my Ph.D. thesis invented and developed GraphChi (, a disk-based large-scale graph computation system which was featured in the MIT Technology Review.

Keywords: data-intensive computation and systems; distributed and parallel computing; graph computation; machine learning and data minining; recommender systems; optimization algorithms.


Software Engineer

– Present (6 months)London, United Kingdom

Facebook acquired startup I co-founded (Moves/ProtoGeo), and I am joining FB as a result.

Plan is to relocate to Facebook in Menlo Park, CA in 2015.

Co-founder (machine learning)

(3 years)

Helped bootstrap the development of Moves app, and advise the company on machine learning, data science and scalability matters.

The company was acquired by Facebook in April 2014.

Ph.D. candidate, Computer Science Dept.

Carnegie Mellon University
(4 years 10 months)

Co-advised by Guy Blelloch and Carlos Guestrin. Working on parallel machine learning framework GraphLab and parallel optimization algorithms. My main research project is GraphChi (, a disk-based large scale graph processing system. Graduated May 2014.

Research intern

(4 months)San Francisco

I worked in the Personalization and Recommendations Team (managed by Pankaj Gupta) and developed and evaluated algorithms for ranking users in the social networks based on specific topics using GraphChi.

Research intern

Microsoft Research Asia
(4 months)Beijing City, China

I worked in the systems research group (mentored by Lidong Zhou and Feng Zhao) to build a system for streaming graph computation (Kineograph). This work resulted in a publication for EuroSys 2011.

Chief Scientist & Co-Founder

Sulake Corporation
(10 years)

For first 4 or so years I was in charge of the technical development and also the lead programmer of the server side of Habbo Hotel. I hired the core of the brilliant development team of Habbo, many of whom are now stars in the industry and founders themselves. I was also part of the management team and participated closely in the funding negotiations. I was part-time 2005-2008 (while studying) and worked mostly in various R&D projects. Member fo the company board 2000-2011.

CEO & Co-founder

Sulake Corporation / Habbo
(1 year 5 months)

Besides bootrapping the technology for Habbo Hotel, I was CEO while waiting to hire a real one.

I founded Sulake with Sampo Karjalainen in May 2000. Sulake's main product was Habbo Hotel (, which had at its peak tens of millions of monthly users and generated up to 50 million euros of annual revenue.

Java Programmer

Satama Interactive
(1 year 6 months)

More or less senior Java programmer on server-side projects.

Lingo programmer

To the Point Oy
(3 years 10 months)

I was lead programmer on numerous cd-rom multimedia titles: corporate presentations, children games, edutainment ..

Honors & Awards

Finnish Engineering Award

Academic Engineers and Architects in Finland (TEK)

Received annual engineering price for founding Habbo (, shared with Sampo Karjalainen.

Finnish Prize

Finnish Ministry of Culture

Cultural price for founding Habbo (, together with Sampo Karjalainen.

We are probably only people in FInland who have received botht the Finnish Prize for arts and the Finnish Engineering Award.

Nokia Foundation Recognition Award

Nokia Foundation

For contributions in collaborative media and social networking (Habbo Hotel). Shared with Sampo Karjalainen.


DrunkardMob: Billions of Random Walks on Just a PC

ACM Recsys

Random walks on graphs are a staple of many ranking and recommendation algorithms. Simulating random walks on a graph which fits in memory is trivial, but massive graphs pose a problem: the latency of following walks across net- work in a cluster or loading nodes from disk on-demand ren- ders basic random walk simulation unbearably inefficient. In this work we propose DrunkardMob1, a new algorithm for simulating hundreds of millions, or even billions, of random walks on massive graphs, on just a single PC or laptop. In- stead of simulating one walk a time it processes millions of them in parallel, in a batch. Based on DrunkardMob and GraphChi [19], we further propose a framework for easily expressing scalable algorithms based on graph walks.

GraphChi: Large-Scale Graph Computation on Just a PC(Link)

OSDI 2012
October 2012

GraphChi can run very large graph computations on just a single machine, by using a novel algorithm for processing the graph from disk (SSD or hard drive). Programs for GraphChi are written in the vertex-centric model, proposed by GraphLab and Google's Pregel. GraphChi runs vertex-centric programs asynchronously (i.e changes written to edges are immediately visible to subsequent computation), and in parallel. GraphChi also supports streaming graph updates and removal of edges from the graph.


Parallel Coordinate Descent for L1-Regularized Loss Minimization(Link)


We propose Shotgun, a parallel coordi- nate descent algorithm for minimizing L1- regularized losses. Though coordinate de- scent seems inherently sequential, we prove convergence bounds for Shotgun which predict near-linear speedups, up to a problem dependent limit. We present a comprehensive empirical study of Shotgun for Lasso and sparse logistic regression. Our theoretical predictions on the potential for parallelism closely match behavior on real data. Shotgun outperforms other published solvers on a range of large problems, proving to be one of the most scalable algorithms for L1.


View Aapo's full profile to...

  • See who you know in common
  • Get introduced
  • Contact Aapo directly

Not the Aapo Kyrölä you're looking for? View more


People Also Viewed

LinkedIn member directory:

  1. a
  2. b
  3. c
  4. d
  5. e
  6. f
  7. g
  8. h
  9. i
  10. j
  11. k
  12. l
  13. m
  14. n
  15. o
  16. p
  17. q
  18. r
  19. s
  20. t
  21. u
  22. v
  23. w
  24. x
  25. y
  26. z
  27. more