skrobot is a Python module for designing, running and tracking Machine Learning experiments / tasks. It is built on top of scikit-learn framework.

Evolutionary Neural Architecture Search framework that improves performance of your DL models

✨ Stock Price Prediction Using Tesla Dataset ✨ In this project, I analyzed Tesla’s historical stock data to forecast future closing prices using machine learning models like Random Forest Regressor. Through data cleaning, feature engineering, and rich visual analytics, I explored patterns in price trends, volatility, and trading volume.

In this project we will try to predict if the person has diabetes has or not.

Machine Learning System Development - Diabetes Prediction

The goal of this project is to build a machine learning model that automatically classifies emails as spam (unwanted) or ham (legitimate). The model will be trained on a labeled dataset and use features extracted from the email content to predict whether an email is spam or not.

Basic iterative hyperparameter search algorithm for custom models and training loops.

This repository performs automated parameter optimization to maximize calcium Ca2+ influx in the trunk of a CA1 pyramidal neuron model while minimizing calcium influx in oblique dendrites, using the Optuna framework for hyperparameter tuning.

On this repo there is a University project about the course of Signal, Image and Video where a constraint was to bulid a project where AI is just a narrow task and not the whole implementation of the project. I will just use AI as a comparison with traditional techniques of video processing.

Built LinkedGen to gain hands-on ML skills; data curation/EDA, fine-tuning DistilGPT2 on Colab, and integrating a working UI with reproducible, deployable workflows. LinkedGen generates professional, tone-controlled LinkedIn posts from user input using a fine-tuned DistilGPT2 model, backed by a clean data pipeline and a Streamlit interface.

Hands-on AI infrastructure from the ground up: GPU memory hierarchy, CUDA kernel optimization, Triton, distributed training, and inference serving. Real benchmarks across the full compute stack, from naive kernels to Groq LPUs, Tenstorrent, AMD MI300X, and Google TPU

This doesn't fine-tune weights of Gemini (not supported by Google), but instead tunes prompt structure — a kind of soft fine-tuning.