Selected work

Systems, products, and AI experiences

A closer look at how BagelTech AI turns complex ideas into elegant, deployable systems.

AI ProductLive product

StructifyAI

Visit site

structifyai.com.au

StructifyAI - Pre-construction intelligence platform

Early EstimateDesign FlawsRevision DeltaFixtures & Fittings

Problem

Builders, estimators, and pre-con teams spend weeks manually reviewing drawings, assembling takeoffs, and researching pricing — leading to costly errors, missed design flaws, and delayed tenders.

Approach

Built a full AI-powered pre-construction intelligence platform that automates takeoffs, drawing review, fixture & fitting schedules, pricing research, and revision checks — so teams can protect their margin before the job starts.

System built

Full-stack AI web application with document ingestion pipeline, vision-based drawing analysis (18+ trade models), early estimate generation, design flaw detection, revision delta tracking, fixture & fittings extraction, and interactive project dashboards. Evidence linked to source drawings.

Stack

Next.jsPythonAnthropic ClaudeGoogle GeminiPostgreSQLVector DBAWS

Outcome

Trade-level estimates in minutes instead of days. Tested on 100+ live tenders with evidence-based results linked to source drawings.

Enterprise AI

Internal Knowledge Copilot

Problem

Large teams struggled with information fragmentation across wikis, docs, and Slack, leading to slow answers and duplicated work.

Approach

Designed a retrieval-augmented generation system with intelligent search, multi-source indexing, and workflow-aware reasoning.

System built

Search and reasoning platform with document ingestion, hybrid search (semantic + keyword), conversational interface, citation system, and access-controlled retrieval.

Stack

ReactPythonLangChainPineconePostgreSQLOpenAI

Outcome

Teams found answers 5x faster with higher accuracy, reducing internal support tickets by 60%.

Agentic Workflow

Multi-Step Automation System

Problem

Document-heavy business operations required manual analysis, research, and downstream coordination across multiple teams.

Approach

Created an agent-based workflow system with multi-step reasoning, tool use, and human-in-the-loop checkpoints for complex operational processes.

System built

Orchestration engine with document analysis agents, web research agents, decision routing, approval workflows, and integration with existing business tools.

Stack

PythonLangGraphOpenAIPostgreSQLRedisREST APIs

Outcome

Automated 70% of manual operational workflows while maintaining quality through strategic human oversight.

Startup MVP

AI-Powered Product Prototype

Problem

A startup needed to validate an AI product concept quickly with a functional, polished prototype that could be shown to investors and early users.

Approach

Rapid product design and development sprint: user research, UX design, core AI integration, and launch-ready front-end in 6 weeks.

System built

Production-quality web application with AI-driven core features, polished UI, onboarding flow, and usage analytics.

Stack

Next.jsTailwindPythonOpenAISupabaseVercel

Outcome

Prototype secured seed funding and onboarded first 50 beta users within 2 weeks of launch.

Have a project in mind?

Let's discuss how we can build something intelligent together.

Start a conversation