Welcome to my personal research and development hub for AI Engineering. This repository serves as a living document of my technical notes, production-ready patterns, and experiments focused on AI Security, Scalability, and Data Integration.
The goal of this lab is to build a foundation for production-ready AI agents and systems, focusing on:
- Security: Defending against prompt injection and system leakage.
- Performance: Optimizing large-file processing and vector search.
- Scalability: Architecting MERN-stack AI integrations that handle enterprise-level loads.
Focus: Protecting AI Agents from hijacking and system prompt leakage.
- Patterns: Sanitizer-Gatekeeper logic, Sandwich defense, and Recency Bias optimization.
- Key Lessons: How to use smaller models to audit larger, more expensive models for security.
- Status: [Active] - Code available in
malicious-prompt-defense/
- Models: OpenRouter, Gemini, OpenAI , Anthropic, Claude.
- Backend: Node.js, Express, Next.js.
- Database: Supabase (PostgreSQL), Pinecone MongoDB.
- AI Tools: LangChain, Vector Embeddings, RAG Pipelines.
Each folder contains: README.md: Explaining the specific problem and the architectural solution.
- Initial Prompt Injection Defense logic.
- Scalable PDF-to-Vector pipeline for Large Files.
- Multi-agent orchestration (Tool Calling).
- Automated LLM evaluations (LLM-as-a-Judge).
Asar Ahmed | AI & Full Stack Developer Building scalable AI solutions and mastering the art of the machine.