An AI engineering studio. Not an agency.

QiYuan Labs is a small studio of senior engineers, focused on building production AI features — and the SaaS and commerce systems they sit inside. We treat every engagement the way internal teams treat their own product.

§ 01 — ORIGIN

Why we started this.

The founding team has spent the last decade shipping production software at venture-backed startups across e-commerce, fintech, and machine-learning infrastructure. In the last two years, every interesting engineering problem we've seen has had an LLM somewhere in the middle of it — and every one of those LLM features has shipped with the same kinds of preventable production failures.

We launched QiYuan Labs LLC in May 2026, registered in Wyoming, to fix that gap. Senior engineers building the unglamorous infrastructure that keeps AI features running — evaluation, retrieval, guardrails, cost telemetry — and the SaaS and commerce systems those features depend on. Honest scopes, written estimates, and a handoff that includes runbooks, monitoring, and the keys.

Today we work with founders shipping AI-native products, e-commerce brands integrating AI into their stack, and operators who need an embedded technical partner without the cost of a full in-house team.

§ 02 — PRINCIPLES

Six rules of the studio.

How we operate, written down. We've been on the other side of consulting engagements that didn't honor any of these — that's why they're principles, not preferences.

01

Honest scoping.

We write a fixed scope before kickoff. If we can't see the path to delivery, we say so — and send you to someone who can.

02

Senior-only delivery.

Every engineer on your project has shipped production systems. No bait-and-switch to junior teams after signature.

03

You own the code.

Work-for-hire. You own the IP, the repo, the deployment, and the keys. Everything we build is yours to keep, fork, or hand off.

04

Ship to measure.

Every project ships with telemetry. We track the metric we promised to move and report against it weekly.

05

Async-first.

Daily written updates in Slack or Linear. Sync time reserved for design reviews, demos, and conversations that need a whiteboard.

06

30-day support.

Every project includes a thirty-day post-launch support window for production bugs and tuning. No surprise invoices.

§ 03 — TEAM

Small. Senior. Direct.

We deliberately stay small. Senior engineers, direct communication, no layers of project management between you and the people writing code.

Stylized portrait illustration of Bo Zeng

Bo Zeng

Founder & Principal

Bo founded QiYuan Labs in 2026 after nearly a decade building production systems at venture-backed e-commerce, SaaS, and ML platform startups. He leads the AI engineering practice — architecture, retrieval, eval — and engagement scoping. Previously: senior engineering roles across ML infrastructure and e-commerce platforms.

QL

Engineering Bench

Senior software engineers

A rotating bench of senior engineers we've worked with for years — specialists in TypeScript, Python, Go, distributed systems, and ML infrastructure. We staff each project with the right people, not the available people.

§ 04 — STACK

What we build with.

Opinionated about tools, pragmatic about choosing them. We pick the boring technology that will still be supported in five years over the latest framework on the front page of Hacker News.

LANG

Languages

TypeScript, Python, Go, Rust (where it matters).

UI

Frontend

Next.js, Remix, Astro, Hydrogen, Tailwind, Radix.

API

Backend

Node.js, FastAPI, Hono, Postgres, Redis, ClickHouse.

OPS

Infrastructure

AWS, GCP, Cloudflare, Vercel, Fly.io, Railway.

AI

AI & ML

Claude API, OpenAI, vLLM, LangGraph, pgvector, Qdrant.

COM

Commerce

Shopify Plus, Stripe, Adyen, Airwallex, Medusa.

§ 05 — OPEN SOURCE

What we publish.

When a tool we've built across enough client projects stabilizes, we open-source it. Below is the studio's public output — small, useful, and maintained.

CLOUDFLARE WORKER · TYPESCRIPT github.com/qiyuanlabs-ai/claude-prompt-cache-edge

A drop-in proxy for the Anthropic API that auto-applies prompt caching.

Sits in front of api.anthropic.com, inspects every Messages request, and adds cache_control markers to the largest stable blocks. Most production workloads see 40–70% cost reduction with no code changes. MIT-licensed.

View on GitHub

More tools coming — eval harnesses, retrieval utilities, and a small library of prompt patterns we keep reusing. Watch the org at github.com/qiyuanlabs-ai.

§ 06 — START

Want to know how we actually work?

Email us or book a thirty-minute discovery call. If we're not the right fit, we'll tell you who is.