Developed an intelligent cologne recommendation platform that uses semantic search and natural language processing (NLP) to match users with fragrances. Users can describe their scent preferences in everyday language (e.g., “I’m looking for something sweet but masculine”).

A Streamlit-based platform offering Ayurvedic remedies. Users can ask queries and complete a questionnaire to discover their dosha (body constitution).

STT-LLM-TTS (websockets/asynchronous) Agent using Deepgram and Groq LPU's and Bert for Vector Embeddings, Chroma for persistent vector db storage and simiarity search for RAG context management

Open Source Hybrid AI Search Application Based OpenAI, Serp, SearXNG, LanceDB...

A project that uses Large Language Models (LLMs) to assist users with legal inquiries, particularly based on the Indian Penal Code (IPC). LegalMind provides case summarization, user-specific legal guidance, and recommendations from past case data.

A prototype for the SIH 2024 Police Department, where users can speak about a crime scenario and receive relevant IPC sections that apply to the situation.

Leadership Coach AI Agent with Youtube knowledge base and Web Search

A test project integrating FAISS vector embeddings with Gemini, deployed on Streamlit Cloud.

An AI-based tool that allows users to upload documents and receive concise summaries and Question-Answering on that Document. The system utilizes vector embeddings to capture key details from the content, enabling quick and accurate response on prompt.

Lightweight RAG system using OpenAI embeddings and ChromaDB for local document search and retrieval.

I post my practices in this repository.

RAG chatbot using car manual data to answer MG ZS warnings via GPT-4o and LangChain.

An internship assignment that creates a FAISS and ChromaDB vector database from a "Luke Skywalker" Wikipedia page for question answering.

A medical bot project implementing retrieval-augmented generation (RAG) for health-related assistance. This bot uses vector embeddings to enhance response accuracy and is designed to provide helpful, real-time information within the medical domain.

'새길' 데이터 파이프라인 & 소비 WAS

A Java implementation of HNSW with multi-vector search support