Loading page...Загружаем страницу...正在加载页面...
AtlasRepo

Sizing a sovereign, air-gapped AI stack for oil and gas health, safety and environment (HSE) in Pakistan

说明与用途

为巴基斯坦石油天然气运营商设计的自主 AI 堆栈,所有数据和模型均在本地运行,提供预警功能,三年成本约 67 万美元,低于云端方案。

问题

An oil and gas operator in Pakistan needs AI that warns before incidents, not reports after, while keeping all data and models within its own infrastructure and avoiding third‑party AI APIs.

结果

The design demonstrates that a sovereign, air‑gapped AI stack can be built on the operator’s own hardware, providing real‑time warnings while keeping data local, with a three‑year cost of about USD 670,000 versus USD 1.85–2.82 million for cloud equivalents.

工作流

  • 1. Select self‑hosted models based on licence to ensure internal ownership.
  • 2. Size the central tier by calculating weight size, KV cache and activations to determine GPU nodes needed.
  • 3. Deploy edge nodes for real‑time detection and forecasting, reusing existing NPU compute.
  • 4. Set up read‑only adapters, Kafka on KRaft, and a unified clock with Chrony and GNSS.
  • 5. Estimate costs and compare to cloud options.
  • 6. Address export licence requirements for 141 GB HBM GPUs.
  • 7. Publish the object model as JSON for reuse.

证据

An oil and gas operator in Pakistan wants one thing from AI in its health, safety and environment department: a warning before the next incident, not a report after it.

来源

None of it may leave the operator's own infrastructure, and no third-party AI API may sit in the serving path.

来源

GLM 5.3 open weights, 753B mixture-of-experts at FP8

来源

Take the largest filed parameter count, multiply by bytes per parameter at the serving precision, then add a planning factor for the KV cache and activations so long incident histories fit

来源

Three years, public prices, electricity at Pakistan's B3 industrial tariff: Own the hardware, with support and power about 670,000

来源

  • 记录日期: 2026-10-03T07:41:08.087Z