Applied AI & LLM engineering — Toronto

Production-grade AI and multi-agent systems, shipped.

Rainbow After Storm is the independent engineering practice of Xun (Mac) Huang — fifteen-plus years of production software, from enterprise platforms at Canadian banks to LLM systems that real users trust.

See the workStart a conversation

What I build

LLM & Multi-Agent Systems
Collaborative agent architectures — planning, structured debate, tool use — orchestrated with LangGraph and LangChain, and designed to be debugged, not just demoed.
RAG & Data Pipelines
Retrieval systems and high-throughput ingestion that feed clean, attributable context into models: embeddings, vector search, streaming pipelines.
Full-Stack AI Products
The whole product around the model — FastAPI backends, React and Next.js frontends, auth, persistence, deployment. AI that ships as software people can actually use.
Strategy & System Rescue
Honest audits of legacy codebases, realistic AI roadmaps, and hands-on technical leadership for teams moving from traditional stacks into applied AI.

Featured work

All case studies →
FinMRI analyzer showing an SPY report with quantitative intelligence panels and the multi-agent progress pipeline
FinMRI — an MRI-style health check for any US stock
A virtual hedge-fund team of AI agents that turns market, news, sentiment and fundamental data into a full-picture risk read — built for understanding what you hold, not chasing signals.
LangGraphLangChainFastAPIRAGPython
AI voice receptionist for live inbound calls
A production voice agent that answers real phone calls, books appointments, and ties into backend systems through webhooks — built on VAPI, Deepgram and ElevenLabs.
VAPIDeepgram STTElevenLabs TTSWebhooks

The engineer behind it

I'm Xun (Mac) Huang. I started in NLP and computer vision when deep learning was just taking off, then spent years building and testing production software for Canadian banks — Scotiabank, TD, and RBC. Regulated environments teach you to build things that hold up.

Today I work almost entirely on LLM and multi-agent systems, and on the part everyone finds hardest: getting AI from prototype to something real users actually trust. My flagship project, FinMRI, is that philosophy running in production.

I build from the underlying tools rather than wrapping a black box — when something breaks in front of a user, you need to know why.

Experience
15+ years shipping production software
Background
Enterprise financial platforms · deep-learning-era NLP & CV
Core stack
Python · FastAPI · React · TypeScript · LangGraph · LangChain · RAG · AWS · Docker · Kubernetes
Based in
Toronto, Canada — remote-friendly

Have an AI system that needs to become real?

Whether it's a scoped build, an integration into existing systems, or senior engineering depth for your team — the conversation starts the same way.

Start a conversation