On a mission to build products developers love, and along the way, teach the next generation of developers. Here's a summary of my work so far.

ScamBuster

I am thrilled to share my recent accomplishment in participating in a highly competitive hackathon. Collaborating with talented individuals, we tackled complex challenges and pushed the boundaries of innovation. Join me as I recount the exhilarating moments and showcase the exceptional projects that emerged from this transformative experience. Read more about my take, and my learnings from the Hackathon here.

• ScamBuster is your ultimate tool for safeguarding against text message scams and spam. In an age where digital communication is paramount, the need to protect yourself from fraudulent messages is more critical than ever. ScamBuster is your trusty companion, armed with cutting-edge AI technology to identify and shield you from potential threats.
• Check out ScamBuster here. Better yet, try it out as well!
• In 2023, I was promoted to VP of Product, now also leading Product Management.

By building ScamBuster, I learnt a lot about language models, machine learning, open source communtiy, Hugging Face models and Android app development. I developed the Android app myself, and worked with a team of 3 to develop the website.

FlowFormer

I recently learnt about Optical Flow, and enjoyed it. I actively delved deeper into the field, and learnt about FlowFormer, the most recent Optical Flow model, which performs highly accurately even for the toughest datasets like the Sintel Dataset.

• FlowFormer is a cutting-edge tool in the realm of time series forecasting. It’s designed to handle sequential data, making it a powerful asset in numerous fields where predicting future events based on past data is crucial.
• What sets FlowFormer apart is its use of transformers, a concept borrowed from the world of artificial intelligence. Transformers excel at identifying patterns and dependencies in data, even when these patterns span across time. This makes FlowFormer particularly adept at dealing with complex, time-dependent data.
• The beauty of FlowFormer lies in its ability to handle the intricacies of time series data, making it a significant advancement in the field. I’ve prepared a comprehensive explanation of FlowFormer, breaking down its logical underpinnings in an accessible manner. I have written a detailed explanation in this pdf.

NeRFs

I learnt about NeRFs, a 3D modelling software, with tons of applications and highly useful features. They take 2D images along with poses as input and give a 3D scene as an output, rendering new views as well!

• I also learnt more about 3D models, more specifically NeRFs, which converts 2D images to 3D models. I gave a general explanation on NeRFs, in an easy-to-understand fashion. Check the detailed explanation I wrote in this pdf!

Handwritten Digit Classification

Coding a CNN (Convolutional Neural Network) was my second project. It classifies handwritten digits to their respective classes, with an accuracy of 99.2%! I loved learning about this model and learnt in-depth about CNN and how to code one. In fact, I even gave a step-by-step explanation of it here!

Function Optimization

Here is my very first project, optimizing complex functions using PyTorch. Leveraging the power of PyTorch, I implemented several optimization algorithms and compared their performance (SGD, Adagrad, etc.) Check out my detailed explanation of the code and optimizing algorithms here!