Selected work from the founding team.

QiYuan Labs LLC was formed in May 2026, but the engineers behind it have been shipping production software for the better part of a decade. The case studies below are drawn from the founding team's portfolio — work delivered at prior startups, on long-running contract engagements, and now under QiYuan Labs. Where clients asked for confidentiality, names are anonymized; the metrics and outcomes are real.

AI · MCP · Open source2026

Front-desk agent tools for Easy!Appointments.

Client. Our own open-source project, built for the clinics, salons and studios that run their bookings on Easy!Appointments, the self-hosted scheduling system.

Challenge. Staff wanted to ask an AI assistant to find a slot, book a patient or move an appointment, without handing it the admin API token or letting it change the calendar on a guess. The booking API is also rate limited, so a chatty assistant could lock the clinic out of its own system.

What we built. A remote MCP server that Claude and other assistants connect to with OAuth. Each person signs in with their own account and picks read-only or front-desk access. Every change is previewed first and only goes through when the person confirms, with an idempotency key so a retry never books twice. A call log masks personal details, and the server caches and caps requests to stay inside the booking system's rate limit. MIT-licensed, with a public demo, and listed in the official MCP Registry.

Read the case study

1 day

Plan → registry listing

Front-desk tools

24/24

Live end-to-end checks

SaaS · B2B2025

Multi-tenant analytics for a logistics SaaS.

Client. A Series A SaaS startup serving freight forwarders across North America and Europe.

Challenge. Customers were exporting CSVs into Excel because the in-product analytics couldn't slice shipment data the way they needed. Churn risk concentrated in the highest-paying tier.

What we built. A customer-facing analytics module with row-level tenant isolation, a custom query builder, and scheduled CSV/PDF report delivery. Backend on ClickHouse with materialized views; frontend in Next.js with a custom D3 chart library.

14 wks

Discovery → launch

3×

Query speed gain

22%

Tier-1 churn drop

AI · Internal Tools2025

Agentic support copilot for a DTC apparel brand.

Client. A direct-to-consumer apparel brand processing eight thousand customer service tickets per month across email and Instagram DMs.

Challenge. Support response time was creeping past twenty-four hours during sales; hiring a third agent was a $60k/year line item the founders didn't want to commit to.

What we built. A Claude-powered triage agent that drafts replies for the support team, with retrieval over Shopify orders, returns history, product specs, and a curated knowledge base. Replies surface in Gorgias for one-click approval. Eval harness runs on every prompt or model change.

62%

Reply-time cut

$11k

Monthly savings

Hallucinations / 90d

Commerce · Shopify2025

Headless Shopify for a beauty brand expanding to APAC.

Client. A US-founded beauty brand launching dedicated storefronts for Japan, Singapore, and Australia.

Challenge. The existing Liquid theme couldn't support multi-currency, localized content, or the brand's new subscription bundle SKUs. Mobile LCP over 3.5s in APAC.

What we built. A headless Next.js storefront on Shopify Hydrogen, deployed to Cloudflare Pages with regional edge caches. Multi-currency pricing via Shopify Markets, localized content in Sanity, custom subscription bundles as a checkout extension.

+18%

Mobile conv. rate

1.4s

LCP, P75, APAC

Markets in 9 wks

AI · RAG2025

Knowledge search for a 60-person services firm.

Client. A consulting firm with twelve years of project archives, contracts, and engagement memos in Google Drive and Notion.

Challenge. New consultants spent the first three weeks of any engagement rediscovering what the firm already knew. Senior partners were the bottleneck on every onboarding.

What we built. Retrieval across Drive and Notion with citations back to source, role-based access matching the firm's existing permissions, and a Claude-powered chat interface. Embeddings refreshed nightly; access decisions evaluated at query time, not index time.

3wk→4d

Ramp time

8.6

User NPS (/10)

100%

Citation accuracy

SaaS · Billing2024

Usage-based billing rebuild for a developer-tools startup.

Client. A seed-stage developer-tools company moving from flat-rate to usage-based pricing.

Challenge. The original billing flow was glued together with Stripe webhooks and a nightly job that double-charged when retries fired. The founder wanted to ship a metered model before the next pricing experiment.

What we built. A metering pipeline on Stripe's usage-based billing primitives, with idempotent ingestion, daily reconciliation against a source-of-truth event log, and a customer-facing usage dashboard. Migrated 1,400 active customers without a single double-charge.

8 wks

End-to-end migration

Billing incidents

+34%

ARPU after change

Advisory · Fractional CTO2024–2025

Fractional CTO for a pre-seed AI startup.

Client. A two-founder AI startup raising their pre-seed round, with strong product vision but no senior engineering leadership.

Challenge. They needed someone to make the architecture and hiring calls that would still hold up after Series A, without committing to a full-time hire before product-market fit.

What we delivered. Six months of fractional CTO support: technical due diligence for investors, the first three engineering hires (sourcing, interviewing, closing), platform and vendor selection, and code review of the founding team's MVP. Handed off cleanly to the first full-time CTO at Series A.

$4.2M

Seed round closed

Senior hires onboarded

6 mo

Engagement length

§ FIELD REPORTS

Word from clients.

People we've shipped for. Names used with permission; some titles abbreviated for confidentiality.

They scoped the project the way an internal team would, not the way an agency does. We knew exactly what we were buying. The analytics module they shipped is now one of our top three reasons customers don't churn.

Marcus R.

Co-founder & CEO, logistics SaaS

Our LLM copilot is in production with zero hallucination incidents in three months. The eval harness they built is now the template for the rest of our AI roadmap. We use them for every AI feature we ship.

Jennifer L.

Head of Operations, DTC apparel brand

They turned a Shopify theme that crashed on Black Friday into a headless storefront handling 10× the traffic at a third of the cost. The team writes code we'd be proud to have written ourselves.

David K.

CEO, beauty brand (APAC)

As fractional CTO during our pre-seed, Bo made calls that would have cost us a year if we'd gotten them wrong. The hires he made for us are still our best engineers.

Sarah C.

Co-founder, AI startup

§ START

Could we ship something for you?

Tell us what you're trying to build. We'll respond within one business day with next steps and a rough estimate — or send you to a partner who's a better fit.

Start a project hello@qiyuanlabs.ai