Data stays inside.
Runs on domestic stack.
Agents still work.
A long-time IT vendor to China's central SOEs, building OA, ERP, and audit systems for energy, finance, and government clients. Every one requires "data inside the enterprise boundary, domestic-stack compliance, fully auditable operations" — which means the vendor's own dev team can't use any SaaS AI. The 500-person R&D + delivery org deployed SiNan privately on 4 domestic GPUs, with domestic OS, database, and CPU throughout — automating the four most time-consuming dev-side activities.
3–5×
Code Review Speed
↓ 50%
New-Hire Ramp
100%
Data Stays Inside
0
SaaS Dependencies
500-person domestic-stack vendor.
Primary clients: central SOEs.
350 R&D + 80 implementation + 40 sales + 30 management. Core business: custom OA / ERP / audit systems for central SOE clients. Customer base: 3 energy central SOEs, 2 financial central SOEs, multiple government units. The vendor itself is SASAC-affiliated (second-tier SOE). All development and office systems must comply with domestic-stack mandates.
SaaS AI forbidden ×
high mentoring cost ×
tightening client demands.
Pain 01
SaaS AI is across-the-board forbidden
SOE compliance rules: data inside the enterprise boundary, domestic-stack compliance, auditable operations. Every commercial SaaS AI physically fails at least one. Developers are stuck working the old way; the efficiency gap keeps widening.
Pain 02
Domestic stack has a steep learning curve
Domestic OS (UOS / Kylin), databases (Dameng / Kingbase), middleware — docs scattered, small communities, senior engineers' "gotcha notes" uncollected. New hires take 3+ months to get productive.
Pain 03
Code review and compliance checks eat senior hours
Domestic-stack projects require strict code compliance: no banned functions, must wire into domestic libraries, must emit full audit logs. Senior engineers spend 30 min/PR on review — the delivery bottleneck.
Pain 04
Client on-site Q&A still needs a senior to fly in
When a client's IT ops hits an issue, they call the senior who originally deployed it. That senior flies in, opens the manual, traces the problem. 5–10 such trips per month — the senior is maxed out.
SiNan privately deployed.
On domestic GPUs.
Data never leaves.
Fully air-gapped deployment: domestic OS + DB + CPU/GPU + domestic LLM (GLM-4.5 on a domestic inference framework). SiNan Gateway wires into GitLab, Confluence, and the firm's compliance rules. 15 domestic-stack-specific Skills shipped; full org uses them via @SiNan in Feishu.
4 domestic GPUs + UOS + Kingbase
Private deployment: 4 inference cards (Huawei Ascend / Cambricon), UOS 1070, KingbaseES v9, domestic Kubernetes distribution. Every machine is domestic-stack certified; the full stack passes SASAC's testing environment.
15 domestic-stack-specific Skills
Dev: domestic-stack code audit, domestic-library invocation docs, adaptation linter, adaptation test generation. Delivery: client-site FAQ, ops manual generator, fault-diagnosis templates, audit report templates. Onboarding: domestic-stack primer Q&A, common-gotchas handbook.
Full GitLab + Confluence integration
SiNan can query the firm's full code repos (permission-scoped) and 8 years of Confluence docs. "Which domestic library for JWT in this domestic-stack project?" · agent returns three prior projects' actual choices + code references.
Upsell the capability back to SOE clients
Once proven internally, the private-agent platform is packaged as a standard deliverable and added to SOE contracts. Two SOE clients have purchased the private version — SiNan experience turned into a new revenue line.
Four workflows,
all on the internal network.
Not a single byte of data leaves the internal network. All inference runs on 4 domestic cards; every call goes into the hash chain.
01 · Domestic-Stack Code Review Assist
DevelopmentBefore
Senior engineer: 30 min per PR · checks banned functions, domestic-library adaptation, audit-log completeness · reviews 10–15 PRs/day max.
After
Agent auto-scans in 5 min · outputs structured checklist (pass / fail / needs-human-judgment) · senior only handles needs-human · 30–50 PRs/day.
Deliverable: PR-review Skill · with domestic-stack compliance ruleset version.
02 · New-Hire Domestic-Stack Onboarding
TrainingBefore
3-month probation · first 2 months spent asking colleagues · same question asked 3–4 times to different seniors · seniors exhausted.
After
"How to configure Kingbase primary-replica under UOS?" · agent returns 3 past-project configs + official doc citations + 2 historical gotchas · new hire contributes solo by week 6.
Deliverable: onboarding from 3 months → 6 weeks.
03 · Remote Support for Client Sites
DeliveryBefore
SOE client's IT ops calls the original implementation senior · senior flies in · opens manuals on-site · 5–10 trips/month · 1–3 days each.
After
Client IT ops @SiNan in Feishu (or the client's own IM) · agent queries the mirrored knowledge base · returns concrete steps in seconds ("use dmrman for backup · check v-sql-port · restart Kingbase") · travel frequency down 40%.
Deliverable: client-support Skill · portable to every new project.
04 · Compliance Docs & Audit Report Generation
ComplianceBefore
Every SOE deliverable needs a domestic-stack adaptation report, security-compliance report, audit-log report · 3 docs × 2 days manual · easy to miss items.
After
Agent pulls project data + compliance rules + audit chain → 3 full reports in 4 hours · PM only reviews final · 100% compliance coverage.
Deliverable: 3 delivery-grade report templates · domestic-stack-cert-ready.
Dev speed up.
Onboarding cut.
New revenue line opened.
3–5×
Code Review Speed
PRs/day from 10–15 → 30–50
↓ 50%
New-Hire Ramp
From 3 months → 6 weeks to solo contribution
↓ 40%
On-Site Support Trips
Client IT ops self-serves remotely
+2
SOE Clients Bought Private Version
SiNan experience converted to new revenue
Full domestic stack: Huawei Ascend + UOS + KingbaseES + GLM-4.5
We're SiNan's customer —
and SiNan is how
we bring AI
to our SOE clients.
— CTO, a domestic-stack systems vendor
Does your team also have to run on domestic stack?
Take a free private-deployment feasibility audit — we'll assess hardware, software, and compliance.
Get Your Free Audit →