AiVA: orders from conversation
Extract order information from business chats and bring it into a structured workflow, with invoices and Kanban-style order tracking.
Back to work
AiVA & CheckBase
Built backend and AI features for business workflows, and co-developed an AI purchasing and inventory app for a retail client.
My roleBackend & AI engineer / Project collaboratorAiVA product page
AiVA project material from my original LinkedIn post.
SourceOverview
Business teams were working across fragmented administrative tools and paper-based inventory logs. These Academy projects explored more useful digital workflows through challenge-based learning.
Project statusAcademy team projects
Scope
Two iOS projects addressing everyday business administration. AiVA explores order management for small businesses; CheckBase focuses on purchasing and inventory workflows for a retail client.
Main features
AiVA: business follow-through
Analytics, calendar and notification features help teams follow orders through the day. An AI assistant supports the broader business workflow.
CheckBase: paper to structured records
Upload purchase orders, extract their contents using AI-powered OCR and map the result into structured JSON with GPT-4o.
CheckBase: purchasing & inventory
A shared workflow connects purchasing records with the needs of finance and warehouse users, replacing paper-based logs.
Technology
Swift & SwiftUI
Native iOS applications and interfaces across the Academy projects.
Apple frameworks
Native platform capabilities, following Apple interface conventions.
Supabase
Database, storage and backend integration in AiVA.
Firebase Messaging
Push-messaging integration present in AiVA.
PDFKit, Charts & WidgetKit
Document handling, visualisation and native extensions in AiVA.
GPT-4o & AI OCR
Structured purchase-order extraction and JSON mapping in CheckBase.
AiVA technologies are supported by its source. CheckBase has a smaller retained artifact set; its extraction flow comes from the project account. No unverified CheckBase database or hosting stack is implied.
Engineering approach
Reduce the number of tools in the workflow
AiVA user testing focused on the overall order-management experience: at most two administrative tools compared with four to five previously.
Use AI at the point of manual re-entry
For CheckBase, purchase-order upload and extraction target the step where paper information otherwise has to be typed into another system.
Build with a multidisciplinary team
These were Academy training projects involving design and engineering collaboration. My contribution centred on backend and AI features for AiVA, and the upload/extraction flow for CheckBase.
What I worked on
- Engineered backend and AI features for AiVA for Business.
- Co-developed CheckBase for a retail client, including the purchase-order upload flow using AI-powered OCR and GPT-4o JSON mapping.
- Collaborated with designers and developers using Swift, SwiftUI, Apple frameworks and the Human Interface Guidelines.
Results & scope
≤2 tools
AiVA order workflow, compared with 4–5
- AiVA user testing showed order-management workflows required at most two administrative tools instead of four to five.
- CheckBase replaced paper-based purchasing and inventory logs.
- Invited by Apple to attend WWDC26 at Apple Park in June 2026.
Built in the Apple Developer Academy training cohort. AiVA’s public website currently offers a waitlist; the project material does not establish a current App Store release.
Project references
AiVA product pageAiVA project updateAnother project
SmartExport
Contact04
Say hello.
For AI engineering roles, collaborations
or a conversation about further study.