llama

Python-based WebSocket for CLI LLaVA inference.

Click below to checkout the website

A comprehensive framework for fine-tuning and evaluating Large Language Models on semantic understanding tasks (WSD & WiC) with LoRA

tsundoku - Using LLM's + Retrieval Augmented Generation (RAG) to help ease the burden of buying books and leaving them unread - even if that means just summarizing chapters and snippets of the book.

Aplicação web para geração automática de conteúdo de marketing otimizado para SEO usando IA (Llama 3.3) e Streamlit

MoSPI web scraper + LLaMA 3 RAG chatbot | Scrapes India's statistics ministry via REST APIs, extracts PDF data, and answers questions using FAISS vector search + local LLaMA | FastAPI · Streamlit · Docker

An AI-powered ATS Resume Analyzer that evaluates resumes against job descriptions — providing match percentage, missing keywords, skill improvement suggestions, and profile summary using LLaMA 3.3 70B via Groq API, built with Python and Streamlit.

A LLaMA-inspired decoder-only transformer using Flash Attention.

Plain-English requests → validated, executable API workflows on a local Llama 3 — zero cloud dependency.

从零预训练 111M Llama:缩小版 Llama 架构 + Wikipedia CLM 训练 + 困惑度评测

Multimodal RAG system — upload PDFs, images, videos & text, then ask questions across all of them.

A 4-stage multi-agent AI pipeline that searches the web, scrapes sources, writes a structured research report, and critiques it — powered by LangChain, Tavily & llama 4.

A hybrid_rag evaluation used as a part of a submission for MSc Data Science Dissertation. Designed to support UK teachers with obtaining access to up-to-date SEND guidance and research.

PM AI Tool is an AI-powered project management system with automatic risk and stage generation, meeting transcription, health checks, weekly digests, and RAG Q&A. Built with Streamlit, SQLite, ChromaDB, Ollama, and Whisper. Runs completely locally

Python trading bot for Binance Futures Testnet — market, limit & stop orders via CLI with live lot-size validation, order management, structured logging, and optional NVIDIA NIM AI analysis.

An AI-powered coding mentor that explains code, detects bugs, generates hints & quizzes, analyzes complexity, and suggests optimizations — powered by NVIDIA NIM (Llama).

Serverless, AI-curated job scraper — aggregates listings from multiple APIs, filters by keywords you configure, and delivers a daily digest to Telegram via GitHub Actions. Works for any role or industry.

TamilTech-QA: A Tamil-English Code-Switched Benchmark for Technical Question Answering with QLoRA Fine-Tuning and Novel Code-Switching Evaluation Metrics (CSPS / TTR / TCF). Published on HuggingFace Hub.

PyQt6 based Llama.cpp Tray Manager for Linux

Semantic code clone classification with Llama LoRA, reproducible preprocessing, inference and preserved QLoRA research.