No description available

No description available

AEGIS (An Edge-AI Guided Inference System) is a multimodal navigation framework designed for human-centric scenarios. It integrates speech recognition, text-to-speech, vision perception, retrieval-augmented generation (RAG), and edge deployment on the SOPHON BM1684X platform, enabling intelligent, real-time assistive navigation.

Aegis is a AI assistant built with Python, Gemini 2.5 Flash and LiteLLM. It handles three domains: grounded electricity bill Q&A using schema-first RAG over parsed PDF bills (ChromaDB & sentence-transformers), weather advice via OpenWeatherMap + LLM recommendation, AC control through a FastAPI mock hardware server.

Agentic Enterprise Graph Intelligence System - Production-ready RAG with Multi-Agent Orchestration

A Retrieval-Augmented Threat Analysis Framework with Meta-Evaluation and Policy-Aware Reasoning

Authorization-first RAG demo showing secure, role-aware retrieval and leakage prevention in multi-agent systems.

AegisLLM is a production-grade LLM system with hybrid retrieval (FAISS + BM25 + RRF), real-time streaming ingestion, and semantic context optimization. It ensures high accuracy (96% Recall@5), reduces token cost by 38%, and enforces safety via multi-layer guardrails and grounding validation.

No description available

A secure AI system that reads complex documents and answers questions using only verified information from those documents, with source citations and confidence scoring.

AegisRAG is a platform where distinct users (tenants) can create isolated "Workspaces." Within a workspace, a user can upload documents, choose an embedding model, select an LLM (Gemini or ChatGPT), and interact with an AI Agent.

AegisRAG is a user-friendly research assistant designed to simplify working with PDFs. Upload, manage, and query your documents through an interactive chat with no technical skills required. Features include PDF-centric insights, smart chat history, and customizable interactions, all in an intuitive interface.

Role-based RAG system that delivers relevant knowledge from documents securely, showing each user only what they’re allowed to access.

APM Portfolio — TrustScore-RAG & Creator Aegis

No description available

Multi-layered privacy framework for RAG systems using NER, differential privacy, and zero-trust architecture. It reduces data leakage by 95% and cuts adversarial attack success to 2%. Built for GDPR/CCPA, it secures AI workloads with minimal accuracy loss.

Local-first RAG: Dagster, Qdrant hybrid, LlamaIndex, FastAPI, Phoenix, W&B Weave, healer

PoC for a secure RAG system on AWS using Terraform and Kendra

🛡️ Aegis-Audit: An Agentic RAG Auditor leveraging Gemini 1.5 and Docling for automated regulatory compliance. Features "Self-Healing" policy logic, clause-by-clause alignment mapping, and persistent multi-session audit memory.

No description available

No description available

No description available

No description available

A local RAG system implemented with Qwen2-7B and FAISS. Features a side-by-side comparison between vanilla LLM and retrieval-augmented generation on a synthetic dataset.

No description available

Backend service for Aegis AI with RAG-aware emotional TTS functionality

이지스-바이오 센티넬스(Aegis-Bio Sentinels): EXAONE-3.5-7.8B와 하이브리드 검색(Naive + BM25), BGE-리랭커를 결합한 의료 전문 RAG 시스템. 검증된 데이터 기반의 지능형 추론으로 할루시네이션을 최소화합니다.

Aegis is a **production-style AI platform** designed to **secure, evaluate, and monitor Retrieval-Augmented Generation (RAG) systems**. It integrates **DevOps, MLOps, and AI Security** into a unified framework.

Aegis OS is an AI Agent Operating System with tool execution, RAG-powered knowledge bases, MCP support, real-time streaming, and multi-agent collaboration, similar to a customizable Copilot Studio for creating intelligent, production-ready AI systems.

Aegis is an Agentic intelligent backend system that leverages Adaptive Corrective Retrieval-Augmented Generation (RAG) to deliver accurate and context-aware answers. It dynamically retrieves knowledge from a vector store, performs web searches, and reformulates queries when needed