Q&A with ASI Alumnus Fellow Khurom - From Astrophysics to Data Science

Fellowship / Getting Started / Data Science / Programming Languages / Machine Learning / ASI / alumni / Careers

Meet Khurom Kiyani, one of our Fellows from our January 2016 ASI Fellowship, as he reflects on his experiences on taking part in the ASI programme. His ASI Fellowship project was partnered with JustEat, where he is now a Data Scientist in their team. Congratulations Khurom!

Tell us about your background, before you decided to move into Data Science?

Before data science, I was a avid academic with an ambition to eventually have my own group. I had nearly a decade's postdoc experience, bouncing between various projects and even disciplines. A physicist by training, most of my career was studying the properties of turbulence, firstly in fluids, and then in the astrophysical scenario, analysing plasma turbulence data from NASA and ESA spacecraft missions. I also worked in signal processing and statistics, computer vision and econometrics, so pretty multidisciplinary.

Eventually, I became disillusioned with the lack of permanent positions in academia and the levels of admin work. When my daughter was born, I looked for a career which would allow me a good level of job stability using the research skills developed in academia. It was a short hop, skip and jump to data science from there.

Why did you choose the ASI Fellowship?

I already had a few data science related interviews - most of which I flopped out of at the second stage. The problem was rarely my technical ability, more my business acumen and talking about the ‘right’ things at interview. The feedback was always that I '... was not quite the right fit for the role'. The fellowship gave me much more confidence at interviews.

Firstly, the fellowship was an opportunity to do a commercial data science project, from start to finish, with a real business focused goal. I chose to do my ASI project with Just-Eat.com, the popular online takeaway company, on a classification problem to customers who are at risk of leaving the use of Just Eat. This gave me immediate confidence at any interview, because I could talk about my industry experience and what I achieved for the business. Having a commercial data science project under your belt is invaluable.

Secondly, ASI provided extensive training in business skills and topics. Every week we made business pitches and had interview and negotiation practice. This transformed my academic presenting style into one industry cares about.

What was the best part of your project or experience on the ASI Data Science Programme?

All of it was pretty good. But I would say my favourite part was the accelerated learning of many new techniques and tech tools, and interacting and making friends with the other ASI fellows, all of whom were incredibly smart. One of the best things about the ASI fellowship is the excellent network it introduces you to.

What would be your advice to people looking to enter Data Science in Industry?

Apply for the ASI fellowship! Seriously, you should do it. Even if you are an expert in machine learning, you will gain a wealth of skills that are hard to achieve in such a short time. I had a whole document full of things that I wanted to have done by the end of the year and after four weeks on the fellowship I had done them all.

The fellowship puts you in a group of very smart people, very kind mentors with their heads screwed on, and an incredible network; all set in a highly dynamic environment.

Beyond this I would advise any prospective data scientist to do five things:

  1. Complete a good course on SQL. Querying databases is essential in data science and SQL is the standard for doing this. With some practice you will begin to appreciate the necessity and beauty of databases and SQL. I would advise starting with codeacademy.com for the easy introduction and to make fast gains; and then moving onto W3schools.com SQL tutorial. Don't worry about the different flavours of SQL just yet - these simple courses will be sufficient to teach you the 80% of SQL that you need to know, like groupby and aggregate operations.

  2. Complete a good course on Python. Alongside R, Python is becoming the language for data scientists. If you are strapped for time like I was, I would complete the excellent codeacdemy.com course. Once you have the rudiments of Python under your belt, I would attempt to translate some of your existing code into Python. I was an extensive MATLAB user, so I did this with my MATLAB functions. As a data scientist you will probably make extensive use of the libraries Pandas and Scikit Learn, as well as plotting with Matplotlib.

  3. Go through a course on machine learning. If you want someone who will make you feel really good, nothing is better than Andrew Ng's Coursera course on machine learning. His style of teaching will make you feel warm and fuzzy inside when you grasp the essential concepts. If you are in a rush, all you need for a decent introduction is listen to his lectures up to 'Logistic Regression'. If like me you really enjoyed this, move onto 'Introduction to Statistical Learning' by James, Witten, Hastie & Tibshirani. This is a wonderful book. Again, if you are strapped for time just read up to chapter 4 on 'Classification'.

  4. Get a data scientist friend who has already made the transition and pelt them with questions. This might actually be the most important thing to do. Many of us who have made the transition were in your position and know how hard it can be to break into industry. We can potentially streamline much of your learning with tried and tested routes and steer you away from dead ends. It's very confusing out there for budding data scientists, so people who have done it already can be very helpful. If you feel comfortable about the jargon, try to attend one or two data science meetups and talk to people there.

  5. Familiarise yourself the tech stack of data scientists and developers as much as possible. You should know how to talk a little about Git version control, coding best practises and the difference between interpreted and compiled languages. Get a little exposure to as many of these as possible. Even better is if you get a Github account, download a free dataset from Kaggle and follow an example data science problem or do one yourself. A data science project is something that you can show to potential employers.

Basically, you should just apply for the ASI fellowship.

How was your overall experience on the Fellowship, and what have you been up to since completing the Fellowship?

The fellowship was probably one of the best decisions I have made in the past 5 years - I can't recommend it enough. Since the fellowship ended, I have interviewed with various companies, finally accepting a data scientist role with Just Eat who I worked with on the ASI fellowship.

Share on: Twitter, Facebook, LinkedIn or Google+