Data and code for FreshLLMs (https://arxiv.org/abs/2310.03214)

The course provides guidance on best practices for prompting and building applications with the powerful open commercial license models of Llama 2.

Python Project Sample for Demonstration

pyWhat LLM version | Answer "What is it?" on the command line with the power of large language models

Dynamic Few-Shot Prompting is a Python package that dynamically selects N samples that are contextually close to the user's task or query from a knowledge base (similar to RAG) to include in the prompt.

A meta-prompting system that transforms raw prompts into production-ready, XML-structured prompts optimized for Claude Opus 4.6. 10 codified rules, 10-component framework, complexity-based routing — based on Anthropic's official best practices.

This project implements a text classification system powered by Large Language Models (LLMs) running locally. The goal is to leverage the capabilities of modern LLMs to automatically categorize and label text data without relying on external APIs or manual human labeling, ensuring privacy, autonomy, and efficiency in text processing tasks.

repo for "An Adapted Few-Shot Prompting Technique Using ChatGPT to Advance Low-Resource Languages Understanding" (2025)

Dynamic Few-Shot Prompting for Customer Support AI Agents A practical implementation of dynamic few-shot prompting using LangChain and HuggingFace models. This repository provides an optimized approach to improving AI agent performance for customer support tasks by selecting relevant examples based on user queries, thus enhancing response accuracy

🛠️ Optimize any raw prompt into a best-practice, production-ready prompt for Claude Opus 4.6 in seconds, enhancing clarity and effectiveness.

[EMNLP 2025] COM-BOM: Bayesian Exemplar Search for Efficiently Exploring the Accuracy-Calibration Pareto Frontier

Leveraged the power of Google Cloud's Vertex AI platform to develop advanced Large Language Models (LLMs). Utilizing the Python API provided by Google Cloud, this endeavor represents a significant stride in the realm of natural language processing and LLMs.

The study explores zero-shot and few-shot prompting strategies using Meta's quantized LLaMA 3.1 70B model to perform Named Entity Recognition (NER) on Nepali text.

Prompt Design & LLM Judge

Engineering Homework solver using ColPali PDF retrieval, Qwen2.5-VL Multimodal analysis, and DeepSeek code generation with schemdraw to create step-by-step solutions with circuit diagrams.

Fine-tuned DeepSeek-Math-7B model to fix viral math memes. Used dataset of 20 errors, few-shot prompting, and Gradio UI on Hugging Face Spaces. Delivered corrections with explanations steps.

A minimal AI chatbot that translates natural language into SQL queries, queries a MySQL database, and displays the results—all via a user-friendly Streamlit interface.

Unlocking the Power of Generative AI: In-Context Learning, Instruction Fine-Tuning, Reinforcement Learning Fine-Tuning, Retrieval Augmented Generation and LangGraph Workflows for AI Agents.

An educational chatbot that employs the Feynman technique—acting as a student and forcing the user to teach a study topic, thereby exposing gaps in the user's understanding.