PipeChat - Civil Engineering AI Assistant

Project Overview

Pipe Chat was an early exploration of applying frontier AI models to civil engineering and water utility management. Built in collaboration with East Bay MUD and UC Berkeley's Center for Smart Infrastructure, it demonstrated the potential for AI-assisted utility operations.

Key Partners

• East Bay Municipal Utility District (East Bay MUD)

• UC Berkeley Center for Smart Infrastructure (CSI)

• SOGA Research Group (Civil Engineering Department)

• Professor Soga and Clifford Chan collaboration

• California water utility professionals

Technical Implementation

• Early GPT-3.5 and GPT-4 integration (3rd-4th LangChain release)

• Retrieval Augmented Generation (RAG) system

• Civil engineering document corpus processing

• Slack bot interface for utility professionals

• Real-time Q&A with engineering documentation

Research Questions

• Can AI safely assist with utility management decisions?

• How accurate are AI responses to technical engineering questions?

• What level of trust can engineers place in AI recommendations?

• How can AI accelerate civil engineering workflows?

Speculative Design - Pipe Core

Developed visionary presentation for AI-powered water utility control:

• Open Compute Project 1U rack design

• Edge inference system for real-time utility control

• Lawrence Berkeley Lab supercomputing integration

• Custom model training for water utility operations

• Real-time command and control AI assistance

Community Impact

Successfully introduced civil engineers to frontier AI capabilities:

• Interactive Slack sessions with California utilities

• Hands-on experience with AI-assisted engineering

• Accelerated adoption of AI tools in civil engineering

• Ongoing research collaborations at UC Berkeley

Long-term Vision

The project established foundations for AI integration in critical infrastructure management, with implications for smart cities, sustainable water systems, and AI-assisted public utility operations.