Inspiration
Electrical installation design under rigorous standards such as NF C 15-100 and IEC 60364 is traditionally a time-consuming, manual, and error-prone engineering process. Electrical engineers and contractors spend countless hours deciphering unstructured architectural notes, manually balancing loads across 3-phase systems, sizing breakers and conductors, and calculating itemized pricing schedules (BPU / Devis). Unbalanced phase loads lead to severe neutral conductor overheating, equipment degradation, and costly project overruns.
We built AutoSchema & BPU Architect to bridge the gap between deterministic electrical engineering principles and autonomous Generative AI reasoning.
What It Does
AutoSchema & BPU Architect is an autonomous AI agent engineered for the Taskmaster Track. It automates the entire end-to-end electrical design workflow:
- Multimodal Document & Note Ingestion: Ingests raw engineer field notes, PDF architectural schedules, and Excel equipment inventories using Google Gemini 3.5 Flash to automatically extract structured circuit schemas.
- Normative Compliance & Audit Engine: Automatically cross-checks circuit configurations against NF C 15-100 / IEC 60364 tables for breaker sizing (In), minimum cable cross-sections ($mm^2$), RCD protection (30mA thresholds), and voltage drop compliance.
- Deterministic Phase Balancing: Implements an optimization algorithm that distributes active loads across L1, L2, and L3 phases, minimizing neutral current ($I_N$) and preventing transformer unbalance.
- Dynamic Single-Line Diagram (SLD) Generator: Renders publication-ready, interactive SVG electrical distribution boards and single-line schematics in real time.
- Automated Bill of Quantities (BPU / Devis): Dynamically calculates accurate material quantities, unit costs, and estimated labor hours for immediate contractor quotation.
How We Built It
- Core Agent Intelligence: Python and Google Gemini API structured with strict JSON schema parsing and chain-of-verification engineering rules.
- Algorithmic Engine: Custom deterministic balancing algorithms minimizing the phase load variance: $$\min \sum_{i \in {L1, L2, L3}} (P_i - P_{avg})^2$$
- Modern Web UI: Responsive interactive dashboard built with Tailwind CSS, Vanilla JS, and real-time SVG vector rendering.
- Cloud Infrastructure: Packaged with Docker and deployed as a scalable, containerized microservice on Google Cloud Run.
Challenges We Faced
- Ensuring Strict Determinism: Generative models can hallucinate numerical calculations. We tackled this by using Gemini strictly for semantic schema extraction while offloading circuit mathematics, breaker curve sizing, and phase balancing to deterministic Python modules.
- Standardization Compliance: Encoding the intricate edge cases of NF C 15-100 (such as dedicated kitchen circuits, socket ring counts, and EV charging points) into reliable verification rules.
Accomplishments That We're Proud Of
- Successfully unifying visual Single-Line Diagrams with instant BPU cost estimations in a single autonomous flow.
- Achieving production deployment on Google Cloud Run with sub-second response times.
- Delivering a truly production-grade tool tailored for real-world electrical contractors and consulting engineers.
What's Next for AutoSchema & BPU Architect
- BIM / Revit Integration: Direct export to IFC and DXF CAD formats for architectural blueprints.
- Photovoltaic & Hybrid Storage Sizing: Adding automated solar inverter, battery bank, and EV charging infrastructure modules.
- Multilingual Localization: Expanding regulatory support for NEC (US) and BS 7671 (UK) standards.
Built With
- docker
- google-cloud-run
- google-gemini
- javascript
- python
- tailwind-css
