
About
One architect. Ten years of shipped systems.
I'm Mai. I started as a freelance developer, building whatever clients needed, and spent years on systems work where software meets the real world: platforms, data services, offline field apps and hardware-adjacent integrations.
As AI became good enough to do real work, that turned into AI architecture: knowledge platforms for enterprise teams, assistants that answer from a firm's own documents, and multi-agent systems that run the admin for a UK property group inside its own cloud.
Mai AI is that work as a studio. One architect, working directly with UK businesses from the first call to the monthly report.
How the studio works
Why one person can deliver what used to need a team
A few years ago, a system like an AI chief of staff needed a project manager, a designer, two or three developers and a tester. Most of that coordination existed to move knowledge between people.
Today I work with AI-assisted engineering: coding agents write and test much of the routine code under my direction, while I own the architecture, review every change and test every path. The design stays in one head, decisions take minutes instead of meetings, and nothing is lost in a hand-off.
The trade-off is honest: I take on a small number of clients at a time, so each one gets the architect, not an account manager.
Values
Four things that do not change
Ownership
The person who designs the system builds it and answers for it. You own the code, the data and the accounts.
Approvals before autonomy
Agents earn trust step by step. Money, publishing and advice wait for a person until you decide otherwise.
Plain language
You will always know what is being built, why, and what it costs. If you do not need AI, I will say so.
Systems that keep running
Monitored, documented and tested on real tasks, so nothing depends on something fragile.
- Model providers
- Anthropic
- OpenAI
- Google Gemini
- xAI
- Groq
- OpenRouter
- AI architecture and strategy
- Agentic system design and orchestration
- Production RAG and GraphRAG pipelines
- Vector indexing and retrieval (pgvector, HNSW)
- LLM integration and evaluation frameworks
- Voice AI infrastructure (VAPI, ElevenLabs)
- Enterprise AI workflow automation
- AI consultancy and technical advisory
- Team training and AI capability uplift
- Full-stack engineering
- TypeScript (strict mode)
- Vue 3 (Composition API)
- React and Next.js
- Nuxt, SSR, ISR and edge rendering
- NestJS
- PWA and offline-first apps
- Zod and state management
- Optimistic UI patterns
- Backend and infrastructure
- Node.js and Go
- Docker (Alpine, multi-stage builds)
- Fly.io, Vercel and Supabase
- PostgreSQL with pgvector
- Redis and BullMQ
- Supabase Edge Functions
- CI/CD pipelines
- Serverless and edge compute
- Security and compliance
- UK GDPR and data residency
- Row Level Security (RLS)
- OAuth 2.0, SSO and SCIM
- SOC 2 and HIPAA-aligned controls
- JWT and biometric authentication
- Audit logging
- Secrets management
- Security testing
- Performance and SEO
- Core Web Vitals (LCP, CLS, INP)
- Lighthouse audits
- Structured data (Schema.org)
- Technical SEO
- CDN and edge caching strategy
- Bundle optimisation
- Sub-100ms response targets
- Image pipeline optimisation
- Generative media workflows
- ComfyUI workflow design
- Stable Diffusion and FLUX
- LoRA fine-tuning
- IP-Adapter and ControlNet
- Image-to-image pipelines
- Automated generation APIs
- AI media production
- Custom node development
Reviews
What clients say
Word for word, from clients in the UK and abroad. Every one rated the work five out of five.
Not sure where to start? Book a Blueprint
A fixed-fee architecture engagement: what to build, how it fits your tools, and a fixed price to build it. £160, credited against the build.