Human | Data Scientist | Writer | Just trying to balance cheese on my pizza

Isn’t writing about yourself difficult and easy at the same time? You only have to research yourself enough to be writing this but that is years of work and difficult in itself. But here I am trying to pen down some aspects of my life and beliefs.

I was never a fond of 160 character long Medium bio in any case. Bio’s are like looking at an unknown person on a street and exchanging a wave. About me’s are like have a coffee or dinner with a stranger. It’s about the story behind the person. …


Stop writing crappy ML Code

When I started with the field of Machine Learning almost 4 years back, this is how my first ML code looked like. I had one file for preprocessing, training, setting hyperparameters and testing.

My code structure from 4 years back

My ‘main file’ was called model1.py and for versioning I had a folder called ‘FINAL_SYSTEM’ which I think had the final code for something! I cant even remember now and to be honest I cant even read my own code. Back in the days ML dev was not that mature and there were many flaws but today that is not the case. …


Not about ‘data privacy’

There are a lot of talks about data privacy where every user’s data needs to be protected and their privacy respected. But what about our learned models? If there is an already deployed model, someone can query it multiple times, collect data and build a supervised learning model. Never thought about this? This indirectly hampers data privacy!

Model Privacy is not an intrinsic quantity associated with a model; instead, it is a measure of information that arises from interactions between the model and its queries.

I was going through last week’s ICLR papers and found this interesting problem which according…


Unsupervised RL and Graph Neural Nets

I attended ICML 2021 which took place last week. Although I am not a big fan of online conferences but then you get to listen to the talks with the comfort of your home. I tried to make notes this time and finally managed to formalize them mostly because such articles have helped me save a lot of time. Specially in the ML field when no one can attend everything that’s happening. Obviously this is not at all an extensive review and mostly of the topics which are of my interest.

Unsupervised learning(UL) for Reinforcement learning(RL)

With both a workshop and a tutorial on UL…


Realizing, unlearning, and respecting

Unspalsh @ Mika Baumeister

Last year during one of the mentoring sessions, someone asked me this question.

Being a women in technology people think I have an advantage at the university, job applications and at all places. Many people around me do not respect me and my skills. They think I get every opportunity just because I am a women. Have you ever faced this? How to handle such situations?

I was taken aback by this question. Most of the questions I was asked were around Machine Learning and Data Science or grad school. This was something I did not expect. …


Practical ways to find your next side ML project

So many fields in ML (Overwhelming?) [Photo from Unsplash- Jan Canty]

There are many resources in the form of online courses for getting acquainted with the field of Machine and Deep learning. I would say there are a lot too many. There are two reasons why taking online courses and doing the projects along with the course might not be as glamourous on your resume-

  1. Thousands of other people are also taking the course and doing the same exercises as you. Online courses are a great way to get to know a skill or study from first principles but eventually you need practical knowledge (if you want to go into industry…


What I am learning and what has been useful

Stole the image from datanami.com

I have been procrastinating learning Reinforcement Learning (RL) since 2 years now. I think the real reason is there’s already so much to learn. So many conferences every other week. Their papers and every field is exciting in its own beautiful way. But here I finally got the chance to block some time off my calendar to learn some RL every week.

My current stage- I have an idea about what RL is (high level definition) and all the buzz words around it ( Like model free RL, exploration vs exploitation). …


Working in ML — Germany, Canada and USA

Photo from Unsplash — windows

After my undergraduate degree I decided to work in a research lab for an year before joining Graduate school. I was working at TU Kaiserslautern, Germany. I also spent a month at the Lulea Technical University, Sweden and around 20 days in Edmonton, Canada attending the Reinforcement learning summer school. As after this I got my Master’s in the US I have some perspective about working as a ML engineer/ data scientist / research assistant in deep learning in 5 different countries (also counting India). I took a different route than most of my friends who were interested in the…


Stop writing crappy models - Unit tests | code formatting

Hey there you are here because you probably think there is a scope of improvement in your coding style. Good code and pipelines not only helps others read your code but you yourself can understand your code after days. In the first part of this series I wrote about how horrible my Keras code looked 4 years back. I also talked about what techniques I am using these days to make it human readable :) where I talked about data versioning, foldering and hyperparameter tuning. …

Purvanshi Mehta

Data Science @Microsoft security research | Editorial Associate Towards Data Science(TDS)

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store