Monday, November 21, 2022

Google Summer of Code (GSoC) 2023

For 2023, GSoC is opening up to allowing anyone (18+) to participate up to two times as a contributor. That means, you can participate as a contributor in 2023 if you have been a contributor once before or never before. But if you have already been a contributor twice, you cannot participate in 2023.
 
Here is an introductory presentation I made for GSoC 2023, updating my previous presentations on GSoC.

Saturday, November 12, 2022

5 years...

Emory 5 years, November 19, 2022
So I am 5 years with the Department of Biomedical Informatics, Emory University this month. However, that is counting in the 7 months I spent in 2016, March - October. If not for that, my continuous 5 years would be on June 19, 2023. In both cases, I am reaching the 5 years soon. This is very long, especially for me as I consider myself nomadic with frequent migrations. This period also includes my 3.5 years of postdoc life since I defended my PhD thesis in 2019 July (ULisboa/Portugal) and 2019 August (UCLouvain/Belgium). The pandemic, travel restrictions, pandemic-related challenges, and other factors made the time go faster, I guess? The years 2020 - 2022 feel like a big monolith, rather than 3 whole years. It is not to say all the three years were same. But they kinda went fast. Really fast. Despite the 5 long years (or 4.5 years, if I do not include the 2016 stint), I was also working with different PIs over my time here. So it was not like I was working on the same projects or topic/domain for the whole time.

In addition to this Emory 5 years, this year feels special to me for two more reasons along the memory lane. First one is, it is 10 years since I left Sri Lanka for my grad school. Then it is also 3 decades since I turned 5. I started counting decades from 15, considering my earliest memories are from when I was 5 years old. So, this year, as I am 35, makes me 3 decades of memories (1992 - 2002, 2002 - 2012, 2012 - 2022). Surely, the last decade (2012 - 2022) was the most eventful and remarkable. I hope the next decade 2022 - 2032 will bring me interesting memories too!

I was interacting with my mentees who apply for gradschool this year. That made me recall my time applying for gradschool in autumn 2011 and my first semester in ULisboa/Portugal in autumn 2012. We had a cool bunch of Erasmus Mundus EMDC MSc 2012 - 2014 group. Our EMDC batch was the middle one among the 5 years program. We had 2 years of seniors before us, and 2 years of juniors after us. That left me in a great spot to observe the past and future of EMDC masters program as I did that. (Same situation with my Erasmus Mundus Ph.D., EMJD-DC as well.) Some of us did a PhD. Others joined the industry after the MSc. Among those who left academia, some started companies. Most others joined big companies in Europe and elsewhere. A few others joined some crazy startups. Some of us left Europe and ended up in third countries (i.e., countries other than our home country and countries in the EU). Some stayed in Europe. Others went back to their home country. Among those who did a PhD, some did an Erasmus Mundus PhD (like me), while others did some other PhD program. Again, after PhD, some joined the industry while the others continued with academia. Among those who continued with academia, the lucky/smart ones became a tenure-track professor instantly, while the others like me became a postdoc. Now, I have the capability to reach out to everyone to give the exact numbers for these categories. But I think that is not important as this is just for my personal blog after all.
Neurasmus page showing students and alumni

EMDC first years were either in Portugal or Spain, where we all spent the 3rd semester in KTH/Sweden. For the forth semester, some of us came back to our host countries (Portugal or Spain), some stayed in Sweden (mostly just with KTH, while some went on to do an industrial internship at Sportify), and some did an internship in a third European country (such as France, Germany, or Switzerland). I came back to ULisboa/Portugal, making it ULisboa/Portugal -> KTH/Sweden -> ULisboa/Portugal for my MSc. That also means, I was one of the few with minimal diverse experiences during my MSc, as I spent time in just two countries where many of us were in three countries. I compensated for this lack of migrations during my Ph.D., with my 3 international research internships (Croatia, USA, and Saudi Arabia) that spanned continents, in addition to my two universities (ULisboa/Portugal and UCLouvain/Belgium). I started my affiliation with ULisboa in 2012 August and spent 7 years doing my MSc and Ph.D., making it the longest university of my life so far, even longer than my 5 years (so far) at Emory University. #WeAreTécnico!

Today a Twitter friend introduced me to Neurasmus, a EU joint-degree MSc program influenced by Erasmus Mundus, but with a neuroscience focus. They also have a nice web page that documents all the students and alumni of the programs in a single place neatly! I wish our Erasmus Mundus programs had something similar. We actually have one for our Ph.D. (EMJD-DC), although not as fancy. But we do not have one for our MSc (EMDC). I remember some of us from the EMDC 2012 intake made a pact to visit Lisboa in 2022 August/September (marking the 10 years). But a pandemic and 10 years have changed many things.
https://twitter.com/pradeeban/status/1580967108321587203?s=20&t=RegDHN3nybldtJxjMNPwzQhttps://twitter.com/pradeeban/status/1580967108321587203?s=20&t=RegDHN3nybldtJxjMNPwzQhttps://twitter.com/pradeeban/status/1580967108321587203?s=20&t=RegDHN3nybldtJxjMNPwzQhttps://twitter.com/pradeeban/status/1580967108321587203?s=20&t=RegDHN3nybldtJxjMNPwzQhttps://twitter.com/pradeeban/status/1580967108321587203?s=20&t=RegDHN3nybldtJxjMNPwzQ

Thursday, November 10, 2022

Docker: failed to create LLB definition

I was getting this weird error on Docker for the same container that was building fine just minutes go.

 $ docker build -t hummingbird:1.0.0 .
[+] Building 0.3s (3/3) FINISHED
 => [internal] load build definition from Dockerfile                                              0.0s
 => => transferring dockerfile: 37B                                                               0.0s
 => [internal] load .dockerignore                                                                 0.0s
 => => transferring context: 2B                                                                   0.0s
 => ERROR [internal] load metadata for docker.io/cellprofiler/cellprofiler:4.0.6                  0.0s
------
 > [internal] load metadata for docker.io/cellprofiler/cellprofiler:4.0.6:
------
failed to solve with frontend dockerfile.v0: failed to create LLB definition: failed to do request: Head "https://registry-1.docker.io/v2/cellprofiler/cellprofiler/manifests/4.0.6": proxyconnect tcp: dial tcp 192.168.65.1:3128: connect: network is unreachable

Turned out, this error can appear when the disk is full (observed in MacOS), with no issue from the built container image itself. How did the desk get full within a few minutes? Of course, by building a few Docker containers and them remaining in cache. Deleting some files to get some storage space and restarting Docker made the build succeed again.

Friday, September 30, 2022

#WeAreTécnico

Exactly 10 years after I joined Técnico Lisboa for the first time for my masters, this month Técnico created an alumni network to connect past and present students and faculty of the university. It has a LinkedIn integration. If you too are from Técnico, feel free to say hi!

Técnico alumni network

Thursday, September 15, 2022

YouTube fake conversations to deceive the innocent onlookers

I encountered this identical conversation between 3 random people on two different videos in YouTube!
A friendly conversation between three strangers? Totally not.

I decided to dig a bit into this. It was easier to spot the deception than I thought. All three accounts were created on the very same day in 2020 (and magically appeared to make the same conversation twice in summer 2022). We can easily conclude that these three accounts were created by the same person or a computer script.




I looked further and found many similar fake conversations happening around topics of investment and crypto trading under popular YouTube videos. An unsuspecting onlooker could totally fall for these scams! All these fake conversationalists have no videos, very few subscribers, and created on the same day (as above) or within a few days/weeks apart - perhaps by a bot and stolen profile photos. Beware!
A similar, but longer "conversation" - but with better "participants" (with accounts not created on the same day). LOL.