Amit Kumar : Life Is Learning!

Year 2021 Review!


So, finally the time has come to write this one. I was quite proud of myself, when I managed to write the previous year review blog (2020), 5 days before the year ended. Unfortunately, this time I am only 10 months late. I can blame 100 things, but the truth is I have been lazy. Okay, enough of guilt trip, lets get on with the review.

So, at the start of the year 2021, I moved to Manchester for 6 months, thanks to pandemic, there wasn’t much happening in London either, so left for Manchester. Now, why Manchester? I was looking to stay in a city, reasonably far from London. Birmingham was too close to London and Edinburgh was too far, so Manchester made the cut.

Manchester on a winter morning in January

Work

Just like most of 2020, I spent pretty much all of my time working for Quansight. Most of the work I did in 2021 was open source. Some notable ones are:

Sgkit

Sgkit is a toolkit for population and statistical genetics in Python. My work involved implementing pairwise distance between all the samples in genomics analysis. In simple terms the problem reduces to finding pairwise distance between all the row vectors in a 2D Matrix.

More information in the blogpost here. Relevant tweet here

Making GPUs accessible via Array API

The libraries in PyData ecosystem like scipy, scikit-learn, scikit-image are largely limited to single-threaded operations on CPU (apart from some exceptions). Due to the popularity and accessibility of GPU hardware, they have become extremely popular in PyData Ecosystem. Therefore, it is very important to have the ability to harness the power of GPUs in typical computations. Since a typical user in PyData ecosystem is already familiar with the APIs of these libraries, it makes sense to be able to use GPUs without writing hardware specific code by just changing the array backend like NumPy, CuPy. This work achieved this via Array API.

Making GPUs accessible to the PyData Ecosystem via the Array API Standard.

More information in the blogpost here. Relevant tweet here

Personal Work

GitHub Actions on your cloud: Cirun.io

This is the service I launched in March 2021. GitHub Actions is great and it’s my favorite CI/CD, but it has some limitations:

  • Machine size is limited to 2 cores, 7GB RAM
  • No support for special hardware like say GPUs
  • Expensive for private repositories: $0.48 per hour

I faced these problems myself at work and came up with an idea to build a service to run GitHub Actions on user’s cloud provider. Fortunately GitHub provides the open source runner software to create self-hosted runners on any machine of choice. Cirun creates on-demand runners on user’s cloud provider.

Here is some money people saved by using Cirun.io

Checkout the marketplace listing here

Personal Life

Personal life hasn’t been very exciting in the first half of 2021, due to pandemic and the new city that I moved into.

A cricket game between Manchester Originals vs Oval Invincibles @ Oval 21st July, 2021

  • Saw some cricket:
    • Eng vs Pak 2nd ODI 10th July 2021 @ Lords
    • The hundred game between Manchester Originals vs Oval Invincibles @ Oval 21st July, 2021
    • Ind vs Eng 1st Test Day 4, 7th Aug 2021 @ Trent Bridge
    • Ind vs Eng 2nd Test Day 3, 14th Aug 2021 @ Lords
    • Ind vs Eng 4th Test Day 5, 6th Sep 2021 @ The Oval
  • Did Wild camping in Lake District
  • Went for hiking in
    • Chiltern Hills
    • Box Hill
  • Ran 5K for the first time in a Park Run event on a Saturday
  • Stumbled into Arjun Rampal at a Lillywhites store in London.
  • Travelled to Edinburgh, Bengaluru, Delhi, Bihar (India) & Budapest, Eger, Szólád in Hungary
  • Tried Bouldering once, it was quite fun.
  • Found the love of my life ❤️
  • Finally went to the winter wonderland in Hyde Park.
  • Attended a ton of friend’s weddings in December.