Hi I'm Ajay Krishnan! I believe in the fact that our lives are too short compared to the number of "Roads Not Taken". This belief coupled with three years of experience as an Undergraduate Computer Science student, internships and a wonderful side hustle experience has resulted in me analyzing and taking up the best techniques for everything I involve myself in.
I take a practical, data-driven approach towards any responsibiliy given to me while making sure that I give it my best towards my job. I like working on completely unrelated jobs to ensure maximum productivity and a chance to pick up more skills and meet more people along the way. In my freetime, you can find me building Deep Learning models, surfing the internet looking for ideas and papers or at a concert supporting the indie music scene. I also like learning more about languages and cultures and have a bilingual proficiency in Tamil, English, French and Hindi.
CodeFemme provides a platform for women from all walks of life to enter STEM and gain a competitive edge.
Independent research on algorithms capable of learning powerful representations from large amounts of unlabeled data using contrastive methods.
Spearheading research on understanding the latent space representations in Self Supervised Learning methodsfor Computer Vision.
This project uses Deep Learning models to trnaslate a hindi text containing in a given image to its English counterpart. Object Detection, Scene Text Recognition and Text Transliteration has been done to obtain the results. The model can be trained to work for other vernacular langauges too!
Implementation of Standard Rock Paper Scissors enhanced by a Deep Learning Model in the back to detect the user’s hand gestures. CNNs are used to extract feature information and predict user's move.Opponent has potential to be improved using AI algorithms.
A model that uses Deep Neural Networks to classify a given object based on its impact on the environment. Built on Pytorch and using Transfer Learning for feature extraction, an accuracy of 88% was achieved for the given task.
A project which predicts whether or not a flight has been delayed on a particular date and also predict by how long it has been delayed. Data has been collected from multiple sources and preprocessed and merged accordingly to provide maximum information.
Built a user interface to perform transactions using SMS. Built a miniature bank server to handle all incoming requests and a server monitoring tool. This product placed first in Smart Finance Hackathon conducted by SASTRA University conducted in association with Paypal.
Used CNN to build a sign language translator to translate sign language convention. Used OpenCV image pre-processing and feature extraction. This project was dropped initially but is bound for more updates in the coming future!
