Flagship Engineering Case Study

ProBuild AI: AI-Assisted Workflows for Construction

QPOI helped build a production construction-tech platform that turns complex documents, blueprints, and project inputs into structured, reviewable workflows for real users.

Duration

10 Months

Platforms

Web (Angular + .NET)

Tech Stack

Angular, .NET Core, SQL Server

Engagement

End-to-End Architecture

The challenge

Construction teams often rely on dense PDFs, technical drawings, and fragmented manual processes. Interpreting that material by hand is slow, inconsistent, and hard to scale.

Off-the-shelf AI was not enough. The workflow needed to be structured, reviewable, and usable inside a production platform rather than presented as a black-box answer.

What QPOI built

  • Document ingestion and staged analysis workflows
  • AI-assisted extraction and task generation
  • Human review interfaces and production platform delivery across Angular and .NET
The Solution

A multi-stage, reviewable workflow

We designed and delivered a multi-stage document and workflow system that helps construction teams move from dense PDFs and project data to structured, actionable outputs inside a live platform.

01

Document ingestion and analysis

Built workflows to process PDFs and technical project documents in stages rather than relying on a single prompt.

Phase 1: OCR extraction and structural analysis.
Phase 2: Identifying core construction elements and dependencies.
Phase 3: Generating structured task outputs that could feed later workflow steps.

02

Human review and correction

AI outputs needed to be reviewable, not blindly trusted. We added revision flows so users could inspect and adjust results when source material was ambiguous.

User corrections could then carry through the workflow, improving reliability while keeping people in control of the final result.

03

Long-context handling

Construction projects generate large amounts of context, so the workflow needed to stay usable across extended task histories and complex documents.

We introduced background summarisation and state management to help the system preserve important decisions without degrading output quality over time.

Production implementation

Beyond the core document workflow, QPOI also contributed to the wider platform infrastructure needed to make the product usable in day-to-day operations.

Real-Time Infrastructure

SignalR • WebSockets • Dual-Channel Streaming

We implemented real-time progress and response streaming so users could see workflow status, follow long-running operations, and interact with results as they were generated. Two SignalR hubs supported both high-level status broadcasting and token-by-token response streaming.

Frontend performance

Web Workers • Angular 16+ • Reactive State

Heavy client-side work was pushed off the main thread where needed, keeping the platform responsive during document compilation and complex UI operations.

Additional platform work

PostGIS • Google Maps Clustering • RxJS Reactive State

Beyond the core AI workflow, QPOI also contributed to marketplace functionality, reporting, and location-based platform features that supported the wider product.

  • Marketplace tools: Map-based search and filtering for relevant opportunities.
  • Reporting support: Interfaces and data flows that helped users work with generated outputs.
  • Platform breadth: Engineering support across multiple operational areas of the product.

Enterprise Team Architecture

RBAC • Seat Management • Tokenized Onboarding

We helped build the admin and account-management capabilities needed for team-based usage, including structured onboarding and role-based access.

  • Invitation flows: Tokenized onboarding links for workspace setup.
  • Team controls: Role and seat management aligned to how construction companies actually operate.

Complex Financial Transactions

Stripe Connect • Commission Logic • Automated Invoicing

We also contributed to the monetisation layer, including billing logic and transaction handling that supported the commercial side of the platform.

Business impact

01
Reduced manual workflow overhead

Turned document-heavy steps into guided digital processes inside a live platform.

02
Improved reviewability

Gave users structured outputs and revision tooling instead of opaque AI responses.

03
Supported platform growth

Extended the wider product through work on real-time systems, admin capabilities, and monetisation infrastructure.

Technology Stack

.NET Core 8Angular 16+Gemini ProAzure CloudSignalRSQL ServerStripe ConnectRedisWeb WorkersDockerCI/CD Pipelines

Need Engineering, Not Just Code?

We build the core systems that power complex businesses.

Let's Build Something Complex