Case Studies

Problem

Traditional shipping workflows require manual package measurement—slow, error-prone, and not scalable.

Solution

• Automated dimension detection from photos using OpenCV (edge & contour analysis, bounding geometry).
• Volume calculation in cubic feet with dual-unit support (inches / ft³ for US shipping standards).
• REST API + web UI with drag-and-drop upload and instant results (< 2 seconds typical).
• Optional archival to Google Cloud Storage; CLI for batch processing.

Stack

• Python 3.9, Flask 2.3, OpenCV 4.8, NumPy, Pillow, Gunicorn
• Docker containers; hosted on Railway
• Optional: Google Cloud Storage for image retention

Impact

How it works

• Preprocess (grayscale, Gaussian blur) → Canny edges → largest external contour → polygon approx.
• Pixel dimensions → scale calibration → cm → inches → cubic feet conversion.
• Returns JSON with dimensions/volume; UI renders a clean result panel.

Architecture

• app.py (API + web), package_scanner.py (CV engine), cli.py (batch)
• Dockerized; Gunicorn workers; health checks
• Env-driven config; structured logging & error handling

Real-world

Built for ROBBINS INTERNATIONAL SHIPPING (RIS) LLC. Production-ready with monitoring and CI/CD via GitHub → Railway.

Links

• Request a demo / integration

Problem

Teams dig through folders and emails to answer repeat questions. Response time is slow, knowledge is siloed, and answers aren’t consistent.

Solution

• RAG pipeline on top of your PDFs, Excel, and Word files (SharePoint/Drive/CSV).
• Private chatbot with source citations and role-based access.
• Embed on WordPress; push escalations to email/Teams.

Stack

• OpenAI + vector DB for retrieval
• Azure/WordPress hosting
• Power Automate for sync and alerts

Impact

Problem

Requests arrive by email and get lost. No SLA tracking, no root-cause view, and limited throughput.

Solution

• Auto-ingest emails (.eml preserved), de-dupe by ConversationId + Message-ID.
• Routing, SLA timers, and status transitions in Microsoft Lists.
• Power BI to forecast volume and flag backlog risk.

Stack

• Outlook → Power Automate → SharePoint/Lists
• Power BI dashboards and alerts

Impact

Problem

Stockouts and over-ordering due to manual counts and lagging vendor data.

Solution

• Daily import of POS/vendor files (NDC/UPC normalized).
• Par levels + lead times → reorder points and alerts.
• SKU profitability & dead-stock views.

Stack

• CSV/Excel intake, Power Query, Power BI
• Optional: OpenFDA metadata enrichment*

Impact

Problem

Laptops, badges, and equipment go missing; spreadsheets can’t show chain of custody.

Solution

• Self-serve issuance/return with approvals.
• QR code labels, mobile scan, photo proof.
• Role-based views and full audit log.

Stack

• Power Apps + SharePoint lists
• Power Automate notifications

Impact

Problem

Stores launch but stall: abandoned carts, slow ops, and no clear growth loop.

Solution

• Performance & UX pass on PDP/checkout.
• Post-purchase flows, inventory sync, reporting.
• Data-layer + GA4 for decisions, not guesses.

Stack

• WordPress/Woo, Stripe/PayPal
• ESP/SMS + GA4, server-side tagging

Impact