Asif's Profile

Personal Website developed using FastAPI and SQLite and deployed in AWS with Nginx reverse proxy server. Currently the architecture of the website has provision to support creating users with their profile details and having the portfolios of the users under the same website. I plan to implement this feature in the future and dockerize the application.

Reproduced the research of Wang, Ji (2021) et al. Converted EEG signals to text tokens using pre-trained BART model and used zero-shot classification algorithm to verify the sentiment of the EEG signals. The resulting algorithm performed on par with the original research in most metrics and outperformed it in some.

Constructed a novel architecture for object tracking using FairMOT and GAN and demonstrated that using generator in a separate layer of the architecture makes the tracking performance much worse because the discriminator learns from the layer itself if the data is from fake distribution or original.

Fine-tuned GPT 3.5-turbo 1106 model with prompt engineered data from real game events coupled with live commentaries to generate new realistic commentaries for new events. Also leveraged text-to-speech and translation APIs to emulate Peter Drury's voice for a more realistic feeling in the generated commentary. The translated commentary, along with closed captions are aimed to be used to make the game more accessible.

Used Tensorflow to recognize various objects in an image, and optimized the algorithm by implementing GPU based execution. Developed during my bachelors (2017-2018), the accuracy of the method turned out to be less than 60% for recognizing everyday objects.

Tried hands-on learning approach with Unity3d platform by developing a clone of Flappy Bird game (the chosen name was Flappy Bat) with personalized object models and custom physics. The game, even though developed well, did not have a very aesthetic appeal because of half-baked assets and sounds, and consequently was not released in any game store.But this experience gave me a headway into game development along with experience with game engines, Blender and C# language around the year 2017-2018.

Learned how to use NumPy, Pandas, Seaborn, Matplotlib, Plotly, Scikit-Learn, Machine Learning, Tensorflow, etc with respect to ML Projects.

Certificate received on 2021-05-06

Learned how to build Shallow and Deep Neural Networks using Scikit-learn and PyTorch.

Certificate received on 2020-01-16

Learned how to improve Deep Neural Networks using regularization, dropout, batch normalization, etc. Dove deep into the black box of deep learning to understand how and why it works so well.

Certificate received on 2020-07-15

Hands-on approach towards learning to use PyTorch for practically developing and building deep learning models. This complements my learning in the classroom for classes like Computer Vision, Natural Language Processing and Machine Learning, where I often designed the models, but wanted to get a more hands-on flavor towards working in the field.