From scratch Pytorch implementation of the original transformer paper. Briefly explain how it works through a sentence-to-sentence translation task example with English to Spanish.

Comprehensive and educational analysis of investment strategies using the performance metrics of GAFAM stocks 📈 - Google, Apple, Facebook (now Meta), Amazon, and Microsoft.
It evaluates two distinct approaches. An active strategy utilizing Moving Average Convergence Divergence (MACD) for trading signals and a passive strategy employing dollar-cost averaging (DCA) with the SPDR S&P 500 ETF Trust (SPY) as a benchmark.

Restore images with centrally masked areas using generative models to accurately predict and reconstruct missing parts.
Post-GAN training, the discriminator and generator collaboratively produce high-quality image samples using continuous gradient-based updates to the activation maps until the samples are classified as real by the discriminator.

Google Chrome extension to automatically track and close inactive tabs making browsing experience lightweight.

Developed a Python-based reinforcement learning agent capable of mastering the Lunar Lander game from OpenAI Gym.
I implemented the REINFORCE loss function using categorical cross-entropy weighted by the discounted reward for each observation and balanced the need for exploration (to learn diverse strategies) with exploitation (maximizing reward outcomes).

Data-driven analysis and machine learning to detect customers who are more likely to churn. Used those insights to propose different strategies for the company to investigate for a potential retention program.

This analysis compares the implementation of linear regression models from scratch (utilizing Numpy and Pandas) with those from scikit-learn to predict the optimal selling price for a client’s house.