H

HuggingFace

Academic NLP project focused on multilingual text summarization using Transformer-based deep learning models and natural language processing techniques.

A full-stack web application designed to bridge the gap between complex mathematical expressions and clean, compilable LaTeX code. This project leverages Machine Learning (Hugging Face models) to take raw text or messy equations and transform them into optimized, industry-standard LaTeX.

The system features a ReactJS frontend for real-time editing and a Python/Flask backend. To keep the deployment lightweight and free-tier friendly (Render/Railway), the project utilizes an External API Proxy Pattern for PDF generation, eliminating the need for a 2GB+ local TeX distribution.

BDS LLM Chat - Hugging Face API Wrapper

A Flask-based chat application that serves as a wrapper for the LLaMA 3.2 model from Hugging Face. This application provides a user-friendly web interface and API for interacting with the model.

Features Modern responsive UI with dark/light mode support Persistent chat history in browser localStorage Markdown rendering for model responses REST API compatible with OpenAI-style endpoints Centralized model worker for efficient inference Redis-backed queue for robust request handling Queue-based architecture for handling multiple requests Admin dashboard for user and API key management External API access with API key authentication MongoDB-based persistence for chats, users, and API keys

Terminal and Web client for Text Completion tasks written in Go - various Hugging Face's hosted LLMs

This is a demo of my angular application for grammatical error correction with transformers based on multilingual Bart correction. The model is trained in 4 languages:

Italian English Russian German