Skip to content

为每个团队打造的
AI 自动化

我们相信每个部门都值得拥有专为其打造的 AI 自动化——而非需要数月客制化的通用工具。

Founded 2026. Pre-seed. Bay Area + Taipei. Customer roster intentionally small while we operate every engagement at founder-depth.

§1 — Founder

Shyan-Ming Perng — University of Chicago, 15+ years of production systems.

Background in physics + computer science (University of Chicago). 15+ years architecting production systems — first UI engineer at PernixData; streaming UI across PS3, iPad, iPhone, AppleTV at Netflix during the DVD-to-streaming pivot; founding engineer of the Nutanix IoT/AI/PaaS platform (Xi IoT) as Principal Engineer; co-founder + Chief Architect at Tiyaro across three platform iterations (pre-ChatGPT democratize-AI, post-ChatGPT enterprise RAG + agents, browser-automation infrastructure).

The pattern across four years at Tiyaro was unmistakable: every enterprise was deploying AI faster than they could govern it. Engineering teams built 80% of the stack themselves, then stalled at the production-grade last 20%. Vendors sold tools and walked away. Consultants delivered frameworks and walked away. Nobody operated the work. JieGou exists to be the third option.

Apr 2026 – present
Founder & CEO, JieGou
Building the AI Operations Partner for engineering-led mid-market IT. Architected the full platform — SvelteKit 2 / Svelte 5 / TypeScript strict / Firebase / Redis / multi-LLM via Vercel AI SDK / EKS deployment. Built Shadow Mode supervision, 10-layer governance, 13 messaging channels, 600+ workflow templates. SOC 2 in flight.
Jul 2021 – Apr 2026
Co-Founder & Chief Architect, Tiyaro
AI platform company through three iterations: pre-ChatGPT "democratize AI" on AWS GPU clusters; post-ChatGPT enterprise RAG + agent platform for GE Healthcare, VMware, HPE, Dell, Cisco; browser-automation infrastructure that directly seeded JieGou. Lived the AI platform transition from both sides.
Sep 2016 – Aug 2021
Principal Engineer / Senior Staff Engineer, Nutanix
Founding engineer of Nutanix IoT/AI/PaaS platform (Xi IoT). Chief architect of the management plane — TypeScript, Golang, PostgreSQL, WebSocket real-time. Championed TypeScript + React adoption across the UI organization. Taught a two-day React course to the engineering org.
Jul 2012 – Sep 2016
Technical Director – UI Architect, PernixData
First UI engineer at PernixData. Oversaw the whole UI architecture — FVP 1.0 / 1.5 / 3.0 across HTML5/JS, Flex web client, AngularJS. Designed Thrift RPC → TypeScript code-generation glue used across the platform.
Sep 2009 – Jul 2012
Senior Software Engineer, Netflix
Built streaming UI across PS3, iPad, iPhone, AppleTV, TV devices during the DVD-to-streaming transition. On-device signup JS library used across mobile / TV / PS3.
Prior
Digeo (Java + Adobe AIR/Flex), Nokia Siemens Networks
Earlier roles building media frameworks, DTCP/IP link protection, UPnP/HTTP streaming, transport control. Deep systems work.
Full profile: linkedin.com/in/smperng

§2 — What JieGou means

Structure for enterprise AI adoption.

JieGou (結構) means "structure" in Chinese. The name reflects what we do — provide structure for enterprise AI adoption.

AI without structure produces unmeasurable outcomes: drafts that nobody approved, decisions nobody can replay, exceptions that nobody catches, audit trails that don't exist. AI with the right operational structure — governance, audit, oversight, escalation, observability — becomes board-defensible.

The structure is what turns AI into operations.

§3 — What we believe

Operator-grade values, not aspirational marketing.

Operate, don't just ship
Selling a tool is easier than running operations. We chose the harder path because the value compounds for our customers — quarter-over-quarter intelligence about their environment that no vendor or consultant can match.
Architect-peer transparency
We don't ask CIOs for trust. We ask them to review the architecture. The /reference-architecture page is what we'd show a peer engineering team before procurement gets involved. If your engineering lead spots a gap, that's the discovery call we want.
Honest about scale
We're pre-seed. Our customer roster is small enough that founder is operationally responsible for early pilots. We'll tell you the math when you outgrow the Operations Partner shape — typically at 5+ in-production pipelines, hiring a platform engineer makes more sense. We'd rather lose the renewal honestly than keep a customer trapped.
Simple beats clever
Background in physics and computer science teaches the same lesson from different directions: elegant solutions are simple, and simple solutions don't fail in surprising ways. Our architecture choices favor primitives over abstractions, deterministic over LLM-based where determinism is required, audit trails over "trust the model."

§4 — What we don't do

Boundary discipline. Things we explicitly will not do, even if asked.

  • ✕ We don't bill hourly. Pricing is engagement fee + annual ops support. Hourly billing creates the wrong incentive for an operations partner.
  • ✕ We don't train ML models on your data. Anthropic, OpenAI, and Google contractually commit to this; we contractually commit on top.
  • ✕ We don't sell tools you'll operate yourself. If your team wants to operate AI workflows in-house, hire a platform engineer — we'll tell you when the math says this.
  • ✕ We don't lock you in. Multi-year terms get pricing discounts but no exit penalty. Customer-VPC deployment means your data + audit trail + outputs all stay with you.
  • ✕ We don't promise "AI transformation." Time-saved math, defensible against measured baselines. If we can't prove ROI after 90 days, you walk away. No exit penalty.

About JieGou

建立您的第一个自动化

看看 JieGou 能为您的团队做什么。

免费开始使用 联系我们