Loading

Aman Patkar | AIcrew Stories

By  aryankargwal

 

๐Ÿค๐ŸผAIcrew Stories come back with another story

The AIcrew blog series chronicles the real-life machine learning adventures of our fellow challenge winners. We also explain their approach and give a sneak peek of the winning solution. Hope this article inspires and motivates you to dive into your AI journey!

In this segment of the series, Vrushank from the AIcrowd team meets up with Aman Patkar, the runner-up for the AI Blitzโšก9: Hello NLP. Aman with his mate Kshitij Singh participated in the challenge as Team Data Pirates.


๐ŸงญAmanโ€™s ML Journey

Graduating this year from the Indian Institute of Technology, Jammu in Mechanical Engineering, Aman was always interested in Machine Learning. He took up courses on probabilistic statistics and general mathematics to further his ML journey.

Aman turned the boring early days of lockdown into a learning opportunity. He took up multiple online courses to polish his ML skills. However, those courses by themselves were not very helpful. The turning point was when he took up problem-solving through ML competitions and hackathons. Actively tackling real AI problems coupled with theoretical knowledge boosted his learning.

One of the most interesting accolades Aman has received is from univ.ai. He was able to bag the second rank in their Geoffery Hinton Fellowship Hackathon. The hackathon, bringing him to the spotlight among the recruiters also paved a path to his current role as a Junior Data Scientist at the Skybox Technologies ApS.


๐Ÿ™‹๐Ÿผโ€โ™‚๏ธGetting started with Challenge

While scouting hackathons and challenges to compete in, Aman found himself exploring the AIcrowd. Aman took up AI Blitz 5 and explored various computer vision puzzles that the challenge offered.

Aman received the challenge launch email for AI Blitz 9:Hello NLP and was driven to do well in this challenge!


๐Ÿ‘จ๐Ÿผโ€๐ŸซAmanโ€™s Advice to you

During the course of Blitz 9, Amanโ€™s solutions were often bested by other participants. This competitive streak motivated him to constantly improve his approach and tweak his solutions. The techniques he learned during the competition were also kept on pen-paper for future reference.

It is hard to keep yourself motivated in face of failure. Aman showed great resilience and self-discipline by excelling in multiple personal projects. His constant drive to improve his solutions to be the best is a testament to the fact.


โ™ŸAbout his submission

Before starting the challenge, Aman only knew the basics of Natural Language Processing. During the course of the challenge, with the help of a starter kit and notebooks by other AIcrew members, Aman learned more about the domain. Falakโ€™s notebook on Using Bert for Emotion Detection introduced him to a new concept. Used BERT for De-shuffling Text, Emotion Detection, and Sound Prediction puzzles and saw a jump in his score! Motivated by his win, he continued to improve other solutions.

Ktrain, a lightweight wrapper for the deep learning libraries TensorFlow and Keras, helped Aman a lot with the unfamiliar algorithms. The library includes pre-trained models for text, vision, graphs, and tabular data. This library paired with the starter kit, helped Aman and the team to achieve great results in all the puzzles. Some of whose examples can be seen in their submission for Emotion Detection.

Want to get a deeper understanding of his solution? Look out for the Challenge Unboxed: AI Blitzโšก9.


Interested in learning more about the winners for this challenge? Check out Sean Ben Hurโ€™s story of how he became the community contribution winner at AI Blitz 9!

What stories do you wanna read next? Comment down below or Tweet us @AIcrowdHQ!

Written by

aryankargwal
Follow

Comments

You must login before you can post a comment.