Weijie Tao
Senior Full-Stack Engineer · Paris, France
I architect systems that scale
I turn ideas into products that hold up in production. In the age of AI I run a fleet of agents alongside me to ship faster — without cutting corners.
Weijie Tao
Senior Full-Stack Engineer · Paris, France
I build the parts of a product that have to be right: domain models, contracts, billing, migrations, the audit trail nobody wants until the day they need it. Mostly TypeScript and Node.js, across B2B and consumer. Based in Paris, working in English, French and Chinese.
Education
EPITECH
RNCP 7 · EQF 7 · Paris, France
Expert in Software Engineering
Programme Grande École · Double degree in Artificial Intelligence of Things
Languages
THINGS I'VE SHIPPED
Real projects, real architecture decisions, real results
Stakimo
Lead Full-Stack Engineer · May 2024 – Present
Lead engineer on a B2B SaaS platform for real-estate developers, from MVP to production. Clients include Nexity, France's largest property developer, and SFLP.
- Shipped the MVP in four months, then built out CRM, back office, client space and the other core modules through to production.
- Built the full reservation-to-signature lifecycle on a domain model of 110 business-rule entities across 13 business modules: options, quick and electronic reservations, withdrawals, e-signature and operational reporting, all engineered for high availability.
- Designed the audit-trail architecture and the versioned database-migration platform, now adopted across the team: 1,200+ automated tests, a Playwright suite, and an isolated Postgres branch spun up per pull request.
- Built the multi-dimensional permission matrix, granular down to user × individual property. Bulk grants compress into two database queries, down from 3.2s to under 300ms. Every permission node is editable platform-wide, and any module can plug in or out.
- Runs on Vercel serverless, with AWS SQS queues and an external-API layer keeping third-party integrations stable. Sentry, SonarQube and business-level instrumentation cover the critical paths in production.
Custyle
Full-Stack Architect · May 2025 – May 2026
AI-native custom apparel commerce platform: full ownership of commerce and monetisation.
- Built the complete monetisation pipeline: cart, orders, Stripe subscriptions and a dual-bucket credit model whose monthly reset fits in a single database transaction.
- Led the architecture and solution design of the AI agent layer: AI-generated design suggestions and automated production orchestration.
- Stood up the project's entire AWS infrastructure and delivery tooling: ECS, ECR and auto-scaling groups, with deployment and CI/CD built end to end so every commit rolls smoothly into production.
- Fixed P0 vulnerabilities across CORS, SQL queries, the OAuth redirect and image upload, then hardened the container image and dependency chain.
- Built the shipping-coverage module that auto-maps Printful fulfilment categories and delivery zones, with Sentry instrumentation on the critical business endpoints.
Tauniqo.ai
Software Engineer · Sep 2023 – Jun 2024
First production version of an AI training and hiring platform, with a Python Django backend.
- Worked on the early build and the full AWS production deployment, shipping the first production version: AI course generation, intelligent tutoring, content ingestion, learner reporting and the admin back office.
- Designed the production-grade RAG and vector-search architecture that grounds all generated content in the customer's own material, traceable to its source, now live at Leroy Merlin, Randstad and Manpower.
- Worked with the core team on the Django REST APIs powering the product frontend.
PR Cockpit
Author · Open source (MIT) · 2026 – Present
A local desktop workbench for reviewing pull requests with Claude or Codex.
- Every review runs in its own isolated git worktree: the agent reads the entire source tree without ever touching your working copy.
- Beyond review: dispatches development tasks straight to the core agents, automatically polls and reviews new PRs, and follows up on PR feedback.
- Findings come back as structured items you can filter, track and process in bulk.
- A human gate on everything: no comment reaches GitHub without explicit approval.
- Integrates natively with existing projects and teams, no workflow changes required. Claude and Codex sit behind one interface, with per-project model and review configuration.
HOW I BUILD
It's not about picking trendy tools — it's about knowing why each piece is there
Cloud Infrastructure
Auto-scaling ECS behind ALB, managed databases, and cache layers that don't fall over at 3 AM.
Rendering Strategies
SSR, SSG, CSR — I don't pick the coolest one, I pick the right one for the use case.
Data & Caching
Cache-aside patterns, query tuning, and multi-layer data architecture. Fast where it matters, reliable everywhere.
SEO & Performance
Pre-rendered HTML, CDN delivery, all Core Web Vitals green. Making search engines and users equally happy.
AI / RAG Pipeline
From vector search to agent orchestration — turning AI from a toy demo into a tool that actually gets work done.
Tech Stack
Frontend
Backend
DevOps & Cloud
AI & Data
Testing & Quality
HOW I DELIVER
No cowboy coding — a clear process that gets the right thing built, on time
Delivery Process
From Concept to Production
I run a structured delivery process that combines agile sprints with hard milestones: requirements broken down with the product owner, architecture agreed before code, sprint planning in GitHub Projects, automated CI/CD, and review gates that a change has to pass before it ships.
- 1Discovery & Spec
- 2Architecture Design
- 3Sprint Planning
- 4Development & Review
- 5CI/CD & Testing
- 6Production Deploy
AI-Powered Development
Multi-Agent Orchestration
I use multi-agent AI systems to parallelize the parts of development that are mechanical but easy to skip. Specialized agents handle code review, testing, documentation and security scanning at the same time — collapsing the wait between writing code and knowing whether it is right, with the quality gates kept in place through version-controlled skill configurations.
- 1Spec → Plan → Tasks
- 2Parallel Agent Dispatch
- 3Code + Review + Test
- 4Security Scan
- 5Auto-Deploy
Agent-Driven Leverage
Built to hand over, not to hoard
A coder agent writes features, a reviewer agent argues with it, a tester agent runs the flows, a security agent scans, a docs agent keeps the architecture map current. The point is not replacing engineers — it is removing the wait between writing code and knowing whether it holds. Every agent config lives in the repo, versioned and reviewable like any other code, so the setup can be read, challenged and reused by whoever else is on the team. I rolled this out across the engineering team, where multi-model code review and a searchable decision log became the default way changes got reviewed, and the setup every new joiner is onboarded through.
- 1Define agent roles & skills
- 2Version-control every config in the repo
- 3Orchestrate: code → review → test → scan
- 4Multi-model review on every change
- 5Hand the whole setup to the team
SAY HELLO
Hiring, building something, or just curious about one of the diagrams above — this form lands directly with me. Email works too.
© 2026 tzstudio · Weijie Tao. All rights reserved.