PowerQueue

The Problem: "Panic Hoarding" in Data Center Construction

We’ve been looking closely at how major data centers are being built right now. It’s a massive construction boom, but it’s hitting a major roadblock: electrical equipment (like transformers) now takes nearly three years to get. The real issue we noticed isn't just the supply shortage; it's how the industry manages it. Big companies or contractors are often building multiple data centers at once, but they manage the equipment for each site in total isolation. If Project A is delayed by weather, its transformer sits in a warehouse as "idle capital," costing the company money. Meanwhile, Project B on the same campus is months ahead of schedule and desperately needs that exact transformer. Because Project B doesn't see Project A's slack, they panic and put in a rush order, which just clogs the supply chain further. Everyone hoards, and everyone loses. Our Solution: PowerQueue We are proposing PowerQueue, an intelligent allocation system that gives a company a "single view" of all their ongoing projects and their equipment needs. Instead of project managers hoarding slots in the dark, PowerQueue treats the company’s entire portfolio of equipment slots as a shared pool.

We aren't trying to build a chatbot; we're building a smart assistant that does three specific things:

Capacity Mapping: We’re using a graph database (Neo4j) to map out every project, every manufacturer slot, and every delivery date. This lets the system spot connections that a human project manager would never see.

The Math (OR-Tools): Instead of just guessing, we’re using a constrained-optimization solver (Google OR-Tools) to run the numbers. It calculates: "If I swap this slot from Project A to Project B, how much money in interest and site delay penalties do I save?"

**Engineering Compliance (RAG): **You can't just swap equipment. Electrical gear has to meet specific local grid codes. We're building an LLM agent that scans engineering submittals and spec sheets to verify if a swap is technically safe and compliant, and it flags it for a human engineer to sign off.

Our Goal

We want to move the industry from "reactive hoarding" to "proactive sharing." By syncing equipment delivery windows with the actual construction schedule, we can prevent companies from buying equipment they don't need yet, and get critical hardware to sites that are actually ready to install it.

Methodology

We're treating this as a logistics optimization puzzle. Optimization: We’re using OR-Tools because the problem of matching N projects to M fabrication slots is a classic math optimization problem. It’s more reliable than asking an LLM to "think" of a schedule. LLM Integration: We are using RAG (Retrieval-Augmented Generation) specifically to handle the "boring" but critical compliance work-checking PDF spec sheets against building codes so the AI doesn't propose a dangerous or illegal swap.

Built With

• faiss
• fastapi
• google-or-tools
• langgraph
• llm
• neo4j
• next.js
• python
• tailwind-css
• vectordb