AI-Powered Observability for Building Infrastructure
0 - 1 SaaS Platform for Intelligent Building Infrastructure
UX Design
User Research
UI Design
Product Design
AI-Powered Observability for Building Infrastructure
Context
Overview
Normal is an AI-native observability platform purpose-built for physical infrastructure. Specifically, the operational systems behind buildings, campuses, and facilities. From HVAC and lighting to water and energy systems, Normal translates vast streams of telemetry data into clear, actionable narratives.
Unlike traditional platforms that overwhelm users with raw logs and fragmented dashboards, Normal leverages language models and inference to provide engineers with insights into what’s happening, why it’s happening, and what to do next. By applying modern UX principles to historically overlooked domains like BACnet discovery, Modbus configuration, and object-level diagnostics, Normal helps operational teams work smarter, faster, and with more confidence.
Challenge
The Problem
As a seed-stage startup, Normal had strong engineering and a clear product vision, but lacked in-house UX expertise and faced a tight timeline for a technically complex domain (BACnet/Modbus, dense telemetry). Building engineers were overwhelmed by object trees, configuration steps, and raw logs. AI needed to be explainable, not just clever.
End users (building engineers and systems operators) often dealt with overwhelming data and configuration tasks. From exploring object trees to identifying failing units, the interaction model needed to feel fast, fluid, and helpful without creating friction. Given the AI-driven nature of the platform, designing a thoughtful interface that contextualized intelligence, not just displayed it, was critical.
Goals
The Solution
The primary goal was to design an interface that made highly technical workflows accessible to users with varying levels of digital comfort. Equally important was establishing a scalable design system that could support growth as the product matured. This meant thinking ahead, not just solving for current needs, but laying a solid foundation for the future.
The product needed to:
Surface system health clearly and concisely
Support complex configuration flows without overwhelming users
Guide interaction with AI models in meaningful, confidence-inspiring ways
Ensure a consistent experience across modules (object explorer, templates, discovery, etc.)
Data
Research Insights
Due to time constraints and the early-stage nature of the product, research was primarily conducted with internal SMEs. These individuals brought deep domain knowledge, having worked in the building automation space across multiple roles and contexts.
Insights gathered included:
BACnet discovery is dense and error-prone without a guided journey
Engineers rely on contextual breadcrumbs to avoid getting lost in object trees
Modbus and legacy device configuration require both power and simplicity
Confidence in AI predictions is built through transparency and explainability
The internal team had a firm grasp on the pain points, allowing us to make informed design decisions quickly.
Strategy
Approach
Working solo as the Principal UX Designer, I embedded directly with engineering and product teams. Given the timeline, we took an agile, iterative approach by breaking the experience down into modules and testing usable versions for early feedback.
We began by prioritizing workflows tied to early demos and sales needs. I focused on usability heuristics, UI clarity, and maintaining information density without cognitive overload.
To stay aligned, I conducted stakeholder check-ins, mapped out feature dependencies, and organized workflows into deliverable phases. By approaching each feature as a mini-product, we ensured that even partial builds could provide standalone value while contributing to the whole.
Design
Process
This phase was where everything came together. Research insights, stakeholder priorities, and technical constraints all translated into tangible design output. Given the 0–1 nature of the product, every screen, flow, and interaction was built from scratch. My focus was not only to deliver polished UI but to engineer the user journey in a way that would reduce overwhelm, support discovery, and highlight system intelligence where it mattered most.
This was a 0–1 build, meaning we were crafting every piece of the experience from scratch. My responsibilities included:
Creating foundational flows and layouts
Designing core modules: onboarding, system health, object explorer, equipment, BACnet/Modbus configuration, templates, settings, and more
Crafting modular UI components for extensibility
Maintaining Figma libraries aligned with Normal’s brand
Validation
Testing
Testing was conducted through informal sessions with internal SMEs and early adopters. Feedback loops were rapid and focused—designed to stress-test clarity, speed, and flexibility in key modules.
This validation process allowed us to:
Confirm comprehension across dense configuration flows
Eliminate unnecessary friction
Ensure UI feedback and system states were clear
Support confidence in AI-generated insights
Outcome
Impact
The product moved from raw prototype to demo-ready software with unified IA and reusable patterns. Engineering could extend features without introducing design debt, while sales could demonstrate a clear narrative of value.
Beyond the UI, we built a UX framework that could evolve alongside the company. The design system laid the groundwork for consistency and efficiency, while the modular architecture gave product and dev teams the flexibility to expand without introducing design debt.
This work enabled Normal to:
Demo confidently to potential customers and partners
Launch a clean, intuitive interface that highlighted the power of their AI
Use the design system as a base for future growth
Stand out in a space where clean UX is rare but deeply needed
The collaboration culminated in a fully realized interface that brought Normal's technical capabilities to life in a way that was approachable and intuitive. This was more than just a design handoff; it marked the transformation of a raw engineering prototype into a market-ready pilot.
Appearance
Identity & Branding
Normal preferred to keep visuals as-is; I kept the UI native to their brand's design language, data-forward, and typographically consistent, while introducing tokens/patterns to scale.
Final Thoughts
Reflection
Designing for operators, not just analysts or developers, reminded me how critical accessibility, guidance, and clarity are when workflows are dense and time-sensitive. I’m proud to have helped shape a product that demystifies complexity and offers engineers a clearer, calmer way to work with the systems they depend on.
