Skip to content

Latest commit

 

History

History
216 lines (167 loc) · 7.37 KB

File metadata and controls

216 lines (167 loc) · 7.37 KB

NVIDIA Inception — 總戰略執行計畫

目標

2026 年 9 月 5 日後重新申請 NVIDIA Inception Program,一次通過。

第一次被拒原因

NVIDIA 明確排除 consulting and outsourced development firms。我們的官網看起來像接案公司。

核心策略

NVIDIA 審核員會做三件事:

  1. 看 ultralab.tw → 這是什麼公司?
  2. 點進各產品 → 是真的 AI 產品嗎?
  3. 讀申請表 → 技術棧有用到 GPU 嗎?

我們要讓三個答案都是「Yes, this is an AI product company using NVIDIA GPUs.」


四個產品的 NVIDIA 角色分配

產品 角色 NVIDIA 價值 現況
MindThread 💰 營收證明 SaaS 訂閱 + NVIDIA GPU 本地推論 ⚠️ 缺英文版
AI Agent Fleet 🖥️ GPU 使用證明 RTX 3060 Ti + Ollama + 自訂 7B 模型 ✅ 已完備
UltraProbe 🛡️ 技術深度 19 向量 AI 安全掃描 + Gemini ✅ 已完備
Ultra Advisor 📊 產品多元性 Fintech AI + Gemini Vision ❌ 缺真正的 AI 功能

執行清單(按優先順序)

Phase 1: 補洞(4月)— 最關鍵

1.1 Ultra Advisor 加 AI 功能 ❌→✅

  • F1. AI 財務分析報告:18 個計算器全部加 AI 分析按鈕(Gemini Flash)
  • F2. AI 保單健檢摘要:擴展現有 OCR → 自動生成健檢報告
  • 詳細規格見 NVIDIA_INCEPTION_ADVISOR.md
  • 執行者:financial-planner repo 的 Claude

1.2 Ultra Advisor 改定位 ❌→✅

  • Landing page 改版:Hero 強調 AI Platform
  • 移除所有 consulting 語言
  • 加「How Our AI Works」技術 section
  • 詳細規格見 NVIDIA_INCEPTION_ADVISOR.md

1.3 MindThread 加英文版 ⚠️→✅

  • 至少首頁 + 功能 + 定價
  • 強調 NVIDIA GPU local inference
  • 詳細規格見 NVIDIA_INCEPTION_MINDTHREAD.md

Phase 2: 強化(5-6月)

2.1 ultralab.tw 更新產品描述

Ultra Advisor 的 AI 功能上線後,更新 i18n:

zh-TW:
  aiseo_desc: "為專業顧問打造的 AI 智能理財平台。18 個 AI 驅動分析工具、
  Gemini Vision 保單 OCR、AI 財務建議引擎。SaaS 訂閱制。"
  aiseo_tag1: "Gemini Vision"
  aiseo_tag2: "18 AI 工具"
  aiseo_tag3: "保單 OCR"

en:
  aiseo_desc: "AI financial advisory platform for professionals. 18 AI-powered
  analysis tools, Gemini Vision insurance OCR, AI recommendation engine. SaaS."

2.2 ProductShowcase 更新 Ultra Advisor metric

// 從
{ name: 'Ultra Advisor', metric: 'AI', metricLabel: '智能顧問平台' }
// 改為
{ name: 'Ultra Advisor', metric: '18', metricLabel: 'AI-powered tools' }

2.3 英文版全站審核

  • 確認所有英文描述不含 consulting 語言
  • 確認 FAQ 對 Ultra Advisor 的描述準確

2.4 Ultra Advisor 英文版首頁

  • 至少 Hero + AI 功能 + 定價

Phase 3: 備戰(7-8月)

3.1 Pitch Deck 準備

一頁式 PDF 或線上頁面:

  • Company: Ultra Creation Co., Ltd. → Ultra Lab (AI Product Brand)
  • Products: 4 live AI products across 3 verticals (social, security, fintech)
  • Technology:
    • NVIDIA RTX 3060 Ti local GPU inference
    • Custom 7B model (ultralab:7b)
    • Gemini Vision + Flash
    • Multi-agent orchestration (OpenClaw)
  • Traction:
    • MindThread: 50+ accounts, 3.3M+ views, SaaS revenue
    • UltraProbe: 19 attack vectors, free tier live
    • Agent Fleet: 4 autonomous agents, $0/mo operating cost
    • Ultra Advisor: 18 AI tools, Gemini Vision OCR
  • Why NVIDIA: Local GPU inference is core to our architecture. Scaling from 1x RTX 3060 Ti → multi-GPU cluster for production agent fleet.

3.2 JSON-LD 結構化數據

所有產品網站加 SoftwareApplication schema:

{
  "@type": "SoftwareApplication",
  "name": "Ultra Advisor",
  "applicationCategory": "FinanceApplication",
  "operatingSystem": "Web",
  "offers": { "@type": "Offer", "price": "..." },
  "description": "AI-powered financial advisory platform..."
}

3.3 llms.txt 更新

ultralab.tw/llms.txt 確認包含所有 4 產品的英文描述和 GPU 技術棧。

3.4 累積數據

  • MindThread: 衝 50+ 帳號、顯示 traction
  • Ultra Advisor: 確認有活躍用戶在使用 AI 功能
  • UltraProbe: 顯示掃描次數

申請表關鍵措辭

Company Description(英文)

Ultra Lab is the AI product brand of Ultra Creation Co., Ltd., a Taiwan-based
startup building and operating AI-powered SaaS products. We have four live products:

1. MindThread — AI social media automation SaaS (50+ managed accounts, 3.3M+ views)
2. AI Agent Fleet — Multi-agent platform with NVIDIA RTX GPU local inference
3. UltraProbe — AI security scanner with 19 attack vector penetration testing
4. Ultra Advisor — AI financial advisory platform with 18 AI-powered analysis tools

Our core infrastructure runs on NVIDIA RTX GPUs for local LLM inference using
a custom fine-tuned 7B model (ultralab:7b on Qwen 2.5 base), achieving 13.2 tok/s
on RTX 3060 Ti. We are scaling to multi-GPU for production agent fleet deployment.

How do you use AI/ML?

- Local LLM inference on NVIDIA RTX 3060 Ti (Ollama + custom ultralab:7b model)
- Gemini Vision for document intelligence (insurance policy OCR + analysis)
- Multi-agent orchestration across 5 platforms (OpenClaw framework)
- AI security penetration testing (19 attack vectors, deterministic + LLM)
- AI-powered financial analysis engine (Gemini Flash)

How would NVIDIA Inception help?

We currently run our entire agent fleet on a single RTX 3060 Ti 8GB.
NVIDIA Inception would help us:
1. Scale to multi-GPU inference for production MindThread + Agent Fleet
2. Access NVIDIA enterprise APIs for enhanced Vision/NLP capabilities
3. Leverage cloud credits to handle inference burst capacity
4. Connect with NVIDIA's startup ecosystem for growth

絕對禁止事項

在所有 4 個產品網站上:

  • ❌ NEVER: consulting, agency, outsource, 接案, 代操, 顧問報價
  • ❌ NEVER: NT$/hr, 按時計費, 專案報價
  • ❌ NEVER: 「幫你做」「幫你建」「代為管理」
  • ✅ ALWAYS: platform, SaaS, automated, AI-powered, subscription
  • ✅ ALWAYS: 「自動化平台」「訂閱制」「AI 產品」

成功標準

NVIDIA 審核員打開以下連結:

  • ✅ ultralab.tw → 看到 4 個 AI 產品、NVIDIA GPU、product company
  • ✅ mindthread.tw → 看到 SaaS 平台、英文可讀、GPU inference
  • ✅ ultra-advisor.tw → 看到 AI 平台、18 AI 工具、Gemini Vision
  • ✅ ultralab.tw/probe → 看到 19 向量 AI 安全掃描
  • ✅ ultralab.tw/agent → 看到 NVIDIA GPU Agent 戰情室

五個頁面,每個都喊「AI product」,沒有一個說「consulting」。


時間線總覽

2026/04   Phase 1: 補洞
          ├── Ultra Advisor 加 AI 功能(F1 + F2)
          ├── Ultra Advisor Landing Page 改版
          └── MindThread 英文版

2026/05   Phase 2: 強化
          ├── ultralab.tw 更新產品描述
          ├── Ultra Advisor 英文版
          └── 全站 consulting 語言掃除

2026/06-08 Phase 3: 備戰
          ├── Pitch Deck 完成
          ├── JSON-LD + llms.txt
          ├── 數據累積
          └── 模擬審核

2026/09   申請
          └── 一次過

最後更新:2026-03-10 狀態:Phase 1 待啟動 所有子計畫:NVIDIA_INCEPTION_ADVISOR.md / NVIDIA_INCEPTION_MINDTHREAD.md