Portfolio RAG Projects
AI Systems Portfolio

11 Enterprise RAG Projects

Production-grade Retrieval-Augmented Generation architectures. Featuring multi-modal parsing, visual troubleshooting, agentic reasoning, and knowledge graphs.

Adaptive Hybrid RAG Layer Icon

Adaptive Hybrid RAG Layer

Advanced Retrieval & Intent Routing

A dynamic routing system that analyzes user query intent to automatically blend or switch between semantic dense vector search and traditional keyword sparse search (BM25).

Tech Stack

Qdrant Milvus Elasticsearch Cohere Reranker FastAPI
View Blueprint Download GUI
Multimodal Enterprise Document RAG Icon

Multimodal Enterprise Document RAG

Complex Data Ingestion

An end-to-end ingestion and retrieval engine designed for complex PDFs, financial reports, and blueprints containing embedded charts, tables, and images.

Tech Stack

LlamaParse ColPali Multi-vector Retriever Vision LLMs
View Blueprint ⚙️ Setup Guide
Corrective Retrieval-Augmented Generation (CRAG) Icon

Corrective Retrieval-Augmented Generation (CRAG)

Agentic Orchestration

An agentic self-correcting RAG pipeline that evaluates the quality of retrieved documents before generation. If the information is deemed insufficient or irrelevant, it triggers an autonomous web-search agent to pull live data.

Tech Stack

LangGraph CrewAI Tavily API Local Evaluator LLMs
View Blueprint ⚙️ Setup Guide
Real-Time Streaming Graph-RAG Icon

Real-Time Streaming Graph-RAG

Dynamic Data & Knowledge Graphs

A continuous ingestion RAG system connected to real-time data streams (e.g., live news feeds or system logs) that instantly updates an interconnected knowledge graph and vector space with sub-second latency.

Tech Stack

Apache Kafka Redpanda Neo4j Bytewax Dynamic Embeddings
View Blueprint ⚙️ Setup Guide
Repo-Level Code RAG Icon

Repo-Level Code RAG

Abstract Syntax Trees & Code Generation

An advanced assistant capable of ingesting massive, multi-file code repositories. It maps structural code relationships (classes, inheritance, functions) rather than just treating code as raw text chunks.

Tech Stack

Tree-sitter Neo4j LlamaIndex ChromaDB
View Blueprint ⚙️ Setup Guide
Privacy-First Guardrailed RAG Icon

Privacy-First Guardrailed RAG

Security & Enterprise Compliance

A secure, production-grade RAG pipeline designed for highly regulated industries (e.g., banking or healthcare). It automatically redacts Personally Identifiable Information (PII) pre-embedding and intercepts prompt injections.

Tech Stack

Microsoft Presidio NeMo Guardrails Llama Guard PostgreSQL (pgvector)
View Blueprint ⚙️ Setup Guide
Context Compression & Token Optimization Pipeline Icon

Context Compression & Token Optimization Pipeline

Cost Optimization & Context Window Management

A data reduction pipeline that strips out filler tokens, redundancies, and irrelevant data from retrieved context documents before feeding them to the LLM.

Tech Stack

LLMLingua LongContext LLMs Custom Token Metrics
View Blueprint ⚙️ Setup Guide
Agentic Long-Term Memory RAG Icon

Agentic Long-Term Memory RAG

Stateful Agents & Memory Management

A persistent memory layer for autonomous AI agents that records user preferences, interaction history, and behavioral shifts, updating its vector index dynamically over time.

Tech Stack

Mem0 MemGPT Supabase pgvector
View Blueprint ⚙️ Setup Guide
Hybrid Cloud-Edge RAG System Icon

Hybrid Cloud-Edge RAG System

Distributed Systems & Architecture

A decentralized RAG deployment that splits the computation workload. Sensitive data or fast lookups are queried locally on edge devices using small models, while heavy queries are offloaded to cloud architectures.

Tech Stack

Ollama ONNX Runtime Transformers.js Cloud Vector Databases
View Blueprint ⚙️ Setup Guide
Automated RAG Evaluation & Synthetic Testing Platform Icon

Automated RAG Evaluation & Synthetic Testing Platform

MLOps & Continuous Evaluation

A continuous deployment evaluation tool that automatically generates synthetic test datasets from raw documents and scores any RAG pipeline on enterprise metrics.

Tech Stack

Ragas TruLens Phoenix CI/CD Automation Pipelines
View Blueprint ⚙️ Setup Guide
Multimodal RAG Product Assistant (P.I.T.O.) Icon

Multimodal RAG Product Assistant (P.I.T.O.)

Multimodal & Visual Troubleshooting

A next-generation Enterprise Multimodal RAG system that uses VLM descriptions and visual similarity to troubleshoot technical manuals with text or photo inputs.

Tech Stack

Gemini 3 Flash/Pro OpenAI GPT-5.2 Gemma 3 ChromaDB Python CustomTkinter
View Blueprint Download GUI