sentence-transformers

A self-contained AI security lab demonstrating document poisoning, indirect prompt injection, and data exfiltration in RAG systems. Explores the "helpfulness paradox" across local and frontier LLMs.

AI document assistant with a production RAG pipeline — hybrid BM25 + semantic search, cross-encoder reranking, and streaming answers. Built with FastAPI, Streamlit, ChromaDB, and DeepSeek LLM. Fully Dockerized and deployed on Railway.

🧠 A persistent memory RAG chatbot that never forgets. Uses Phi-3-mini LLM, ChromaDB vector database, and Ebbinghaus memory decay curves.

AI-powered resume and job-description analysis platform with document parsing, skill-gap detection, ATS insights, and structured evaluation.

Explainable agentic job-fit analysis system using LangGraph, semantic matching, and Streamlit.

From Scientific Papers to Knowledge Graphs: Exploring Agentic AI Literature.

Medical RAG chatbot answering diabetic retinopathy questions with cited sources, built with LangChain, ChromaDB, sentence-transformers, and Llama 3.3 70B (Groq). Deployed on HuggingFace Spaces.

🧠 HireGraph — AI-powered multi-signal candidate ranking engine | INDIA.RUNS Hackathon 2026 | Beats keyword filters using semantic embeddings + career trajectory + behavioral intelligence

Built an agentic RAG system (Observe→Plan→Act loop with relevance grading, query rewriting, and citation-grounded answers) using sentence-transformers embeddings, a ChromaDB vector store, and a provider-agnostic LLM layer (Claude API), with an evaluation harness (Hit@k, MRR, faithfulness), a FastAPI service, Docker packaging, and GitHub Actions CI.

RAG pipeline for semantic querying and structured extraction from technical and regulatory PDFs. Built with sentence-transformers, FAISS, Claude, MLflow, and FastAPI.

Production-style RAG chatbot with hybrid retrieval (BM25 + BGE dense + cross-encoder reranking) and parent-child chunking over a 183-page policy PDF. Fully free stack — Groq, ChromaDB, sentence-transformers, Streamlit.

RAG-based Document Q&A chatbot - retrieves answers from documents using Sentence-Transformers + FAISS, grounded via a local LLM (Ollama).

A simple context-aware customer support chatbot built with Python and Streamlit. It answers common customer queries related to refunds, shipping, business hours, and contact information while supporting follow-up questions.

AI-powered career platform — resume scoring, semantic job matching, cover letter generation, interview prep, and application tracking. Built with Claude, Streamlit, and sentence-transformers.

Local-first support agent built with LangGraph + local Hugging Face models (sentence-transformers + Qwen2.5-0.5B-Instruct). No hosted LLM APIs — retrieval, generation, and verification all run on-device.

AI-powered FAQ chatbot with semantic search — understands meaning, not just keywords. Built with Streamlit, sentence-transformers, ChromaDB, and Groq LLM. Features a dynamic, hot-reloadable system prompt, two-sided chat UI, and guideline-driven handling of small talk, off-topic questions, and safety edge cases.

Built an AI-powered PDF chatbot using Python, Streamlit, LangChain, ChromaDB, sentence-transformers, and Google Gemini to implement Retrieval-Augmented Generation (RAG) over uploaded PDF documents.

Fully local RAG chatbot for GIS knowledge — sentence-transformers + ChromaDB + Ollama

AI-powered HR candidate screening and shortlisting system

Chat with your PDFs using RAG — LangChain, ChromaDB, and sentence-transformers with a Streamlit UI. Supports Ollama (local) and Claude API backends.

Using FastAPI, React, MySQL, FAISS, and Llama 3

A FastAPI RAG service for multi-jurisdiction employment-law Q&A with streaming responses and automated evaluation metrics. Supports multiple LLM providers (Anthropic, OpenAI, Ollama) with prompt caching and ChromaDB vector retrieval.