This is an intentionally over-engineered portfolio site showcasing deep AI Product Engineering, product design, and strategic decision-making capabilities, architectural rigor, and end-to-end craftsmanship.
Production AI systems and engineering architectures.
A modular AI marketing automation system built with n8n, Groq, Supabase, Cloudinary and WordPress. The delivered newsletter workflow coordinated content input, AI drafting, media handling and publishing. Wider plans included social automation and retrieval-based content memory. Public details remain subject to client approval.
An AI-native portfolio built as a working demonstration of AI Product Engineering. It combines a Next.js website and Sanity CMS with a planned RAG assistant, custom ML router, LangGraph orchestration, evaluation and personalized voice. The foundation and CMS are in progress, while the AI and voice layers will be added through later verified milestones.
Deep dives on AI architecture, ML routing, and observability.
I am turning my portfolio into a complete AI product rather than placing an AI feature on top of a static website. This first update explains my move from product design into AI Product Engineering, what has been built so far, and the planned path through managed content, RAG, ML routing, agents, observability and voice.
Target design: Custom classifier exported to ONNX serving intent, route, and Roman Urdu predictions in < 5ms without LLM latency.
Target architecture: Qdrant vector index paired with Neon PostgreSQL canonical chunks. 100% rebuildable, versioned, and cited.
Target state engine: Typed graph with conditional retry edges and inspectable execution telemetry for full transparency.
Operational constraint: Engineered strictly within permanent free-tier quotas with automated budget audit runbooks.