Work

Real systems, not concepts.

One of these is built and working. The other is still in progress. Both are real projects with real code — nothing here is a hypothetical case study.

OCR & Document Classification

A system that reads and sorts documents on its own

Public code — verify it yourself View on GitHub →
01 — The Challenge

Manual document sorting doesn't scale

Organizations receive a steady stream of scanned documents and photos of paperwork. Sorting and filing them by hand is slow and error-prone.

02 — The System

OCR plus automatic classification

A system that takes a scanned document — or just a photo of one — extracts the text with OCR, and automatically classifies what type of document it is.

03 — How It Works

Extract, then classify

Text is extracted from the image using optical character recognition, then the extracted content is run through a classifier that determines the document type — no manual sorting step required.

04 — Technology

Computer vision and classification models

Built on OCR and computer vision for text extraction, with a classification layer for document typing.

OCR Computer Vision Classification Models
05 — Status

Built and working

The system reads and classifies documents, including imperfect scans and photos rather than only clean, flatbed-scanned input. Full code is public on GitHub.

Document Intelligence Preview
scan_0142.jpg
InputPhoto of document
OCR ExtractionComplete
Document TypeClassified
StatusSorted — no manual step
Interface demonstration — illustrative field labels, not live data.
In progress · Construction & Risk Management

Risk visibility for construction projects

In development — public repo View on GitHub →
01 — The Challenge

Invoices and project data drift apart

On construction projects, invoices and project records can quietly stop matching — and projects that were expected to win can start losing ground — without anyone noticing until it's expensive.

02 — The System

A dashboard with a chatbot

Currently building: a dashboard, paired with a chatbot, that cross-references invoices against project data for construction companies.

03 — How It Works

Cross-reference and flag

Invoices are checked against project records. Mismatches are flagged, and projects trending away from expectations are surfaced — instead of being found the hard way, later.

04 — Technology

Data cross-referencing and conversational access

Structured data comparison across invoice and project records, with a chatbot layer for querying the dashboard directly.

Data Extraction & Structuring Anomaly Detection Conversational AI
05 — Status

In progress

This system is actively being built. The interface below shows the concept currently in development, not a finished product.

ConstructAI — concept In development
Project A — invoices vs. recordsMatches
Project B — invoices vs. recordsFlagged — mismatch
Project C — win probabilityLosing ground
Concept mockup — in development, values are illustrative placeholders.

Have a problem worth solving?

Let's engineer the intelligence behind it.

Start a Project → See services →