Sub-second
Codexia Live coaching loop target
Designed around live learner events, not delayed chat-style prompts.
AI systems engineer / full stack product builder
I'm Azmi Muwahid, a senior full stack engineer focused on agentic workflows, RAG, tool calling, and real-time interfaces. My work sits where product UX, backend reliability, and model behavior have to line up.
Proof, not hype
Sub-second
Designed around live learner events, not delayed chat-style prompts.
Reviewable
AI summaries enter product workflows with human review states.
Server-only
Notion tokens stay inside SvelteKit load functions and server modules.
Signature interaction
A lightweight map of how I think about agentic product work: signals become context, policy, tool calls, interface feedback, and eventually a measurable product outcome.
Active layer
Prompt, voice, pause, click, or business event enters the system.
The tools I use when model behavior, product UX, backend reliability, and cost controls all matter.
Selected case studies
Each case study documents the problem, constraints, architecture, decisions, and tradeoffs behind the work.
An AI-powered website builder allowing Indonesian small-medium businesses (UMKMs) to construct their business landing pages in 60 seconds.
Successfully built a multi-tenant platform with visual editor capabilities and automated generation pipelines, lowering the digital commerce entry barrier.
A real-time coding coach that watches friction signals and responds with progressive hints instead of handing over answers.
Converted an open-ended AI tutor idea into a low-latency coaching loop with measurable intervention points.
An agentic B2B CRM for the mattress and bedding industry that balances autonomous AI customer interactions with transparent telemetry.
Engineered a next-generation CRM inbox where human operators oversee autonomous lead extraction agents via live Model Context Protocol streams.
An automated serverless SaaS platform converting long-form YouTube videos into viral vertical Shorts, optimized for mixed-language regional audiences.
Delivered a fully automated media rendering pipeline that transcribes, isolates viral hooks, and crops landscape videos into portrait formats.
Whether you want to talk about tool_calls, the future of RAG, or why Ruby on Rails is still a powerhouse for AI backends—reach out.