I am an Artist.
I am an AI Enthusiast.
I am a Developer.
I am a Researcher.
Currently an MSc. student in Informatics at Technical University of Munich, Germany. My interests include Artificial Intelligence, Machine Learning, Computer Vision, Deep Learning and NLP. Currently undergoing an IDP at Convaise UG, Munich where I am working on finetuning SOTA models including ProphetNet,T5 and other models for my usecase.
Have completed various short projects in machine learning/NLP as part of various hackathons and am interested in internships/a research career regarding the same.
I mainly code in Python, C++ and Javascript. Extremely interested in Competitive Coding, with my main interests being in Algorithms and Graph Search Techniques.
Working in ML for more than 3 years, mostly in Supervised Learning. Mostly use Python or Javascript with Tf.js for coding up solutions.
Worked in Computer Vision, Sequence Modeling and Generative Modeling. Mostly Code in Python using Keras, Tensorflow and ScikitLearn. Have implemented many projects regarding the same and completed several courses.
Worked with Matplotlib, Seaborn among others for data visualization. Mostly use Pandas for Data Preprocessing.
As DS&ML Domain Head at Gen-Y I lead a monthly paper presentation on Deep Learning, AI, CV etc and related topics. My main interests lie in Computer Vision and Sequence Modeling. Currently working on projects related to the same. More Here.
Active Blogger on Medium, writing on my varied interests including Literature, Music, Philosophy and Computer Science. See More.
Writing Poetry for more than seven odd years. Mainly write in English and Bangla. Check out some here!
Have been learning the Classical Guitar for 4 years under curriculum from London College of Music. Finished Grade V with Distinction.
A basic recommendation system based on the user's mood using OpenCV and Deep Learning for detecting emotions from the face of the user and then recommending personalized Songs and Movies. The output page also contains the image of the user with Style Transfer applied to it based on the mood detected with paintings of great artists who have portrayed similar emotions through their art.
Based on a stacked LSTM model and data from a US Names Dataset, we generate novel names based on the data. The data is preprocessed using a char2id system and Run through the Deep Learning Model.
Bengali is the 8th most spoken language on earth. I created a ConvNet Model using Keras and trained it on a Bengali Digit Dataset. The Trained model can predict the correct digit with a 95.8% Accuracy.
Implemented Polynomial Regression using Tensoflow.js, working with a 2nd Degree Polynomial and the Adam Optimizer.