AI engineering learning roadmap
A record of personal projects and technical exploration. Implemented features are distinct from validated production readiness.
Project 01 · Vision & Multi-modal
Smart Inspector
Azure OpenAI image analysis with Pillow validation and metadata stripping. Product price/link fields are model-generated, not live shopping results.
Project 02 · Conversational AI & RAG
Support Bot
Foundry agent answers grounded in a bundled FAQ. Returned PII entities are masked; the original question still goes to Azure. No vector search or persistent conversation service.
Project 03 · Prompt Engineering & Content Gen
Prompt Forge
Azure OpenAI generation with Content Safety required in production and a limited development heuristic. Request history is held in browser state.
Project 04 · Knowledge Base & RAG
pocketRag
PDF, Markdown, TXT and DOCX ingestion; SQLite/NumPy retrieval and Azure embeddings/generation. Owner-scoped uploads expire after 30 minutes by default.
Project 05 · Autonomous Agents & Tool Calling
Agentic Workflow
Offline planner or Azure model with calculator, bundled knowledge lookup and approval-gated simulated alerts. Token-protected sessions remain in memory.
Project 06 · AI Agents & MCP
Agent Workspace
Offline or Foundry adapter. Conversations are owner-scoped in memory or optional Cosmos DB; MCP tasks and notes share a workspace store.
Project 07 · Multi-Agent Systems & Orchestration
Brief Generator
Offline fixtures or Azure generation. No web research or independent fact verification; source labels and metrics are fixtures or model output. Brief history is process-local.
Project 08 · AI Deployment, Observability & AIOps
AI Ops & Deployment Studio
In-process lab with offline simulation or Azure calls, chaos scenarios and session metrics. Does not deploy infrastructure or control external production traffic.