Project

Agentic System Design Chatbot (Industrial Water Treatment)

An expert assistant that helps engineers translate requirements into system design decisions, surfaces constraints, and produces clearer documentation — speeding up early-stage design cycles while keeping traceability.

Industrial water treatment with AI dashboard
Requirement intake to structured design notes — consistent early-stage engineering decisions.

Implementation-focused summary

Guided engineering decisions, consistent documentation

Built to support design conversations by asking the right questions, applying domain constraints, and keeping outputs structured and usable.

Problem

Early-stage system design required pulling information from scattered references and repeating similar reasoning across projects. Engineers needed a faster way to iterate on designs while maintaining consistency and a clear record of assumptions and constraints.

Solution

An agentic design assistant that guides the intake of requirements (influent characteristics, targets, footprint, constraints), recommends design considerations, and generates structured design notes and documentation artifacts.

How it works

1

Guidance

Requirement intake and structured questioning to reduce ambiguity.

2

Constraints

Encodes key constraints and tradeoffs to keep proposals realistic.

3

Outputs

Produces clear design notes and documentation-style summaries.

Architecture & Stack

  • Conversation engine with structured state capture (requirements, assumptions).
  • Decision support routines for design trade-offs and constraint checks.
  • Templates for turning conversations into documentation artifacts.
  • Governance patterns to reduce hallucinations and ensure safe recommendations.

Impact

  • Faster early-stage iterations through guided requirement collection.
  • More consistent documentation and clearer rationale for decisions.
  • Reusable system for onboarding and standardizing design discussions.

Your role

Designed the interaction model, implemented structured state capture, and aligned the output format with engineering documentation expectations.