Sanitized architecture case study
From mobile visual search to video commerce at scale
At Olive Theory, I built a React Native shopping experience that let a user photograph a product and discover visually similar items available to buy. That product-matching foundation evolved at Accio into a video platform that surfaced relevant products inside the player when a viewer paused.
This case study deliberately omits proprietary model details, customer data, vendor configuration and internal infrastructure topology. It describes the system boundaries, product decisions and verified operating scale.
Verified outcomes
- mobile downloads in month one
- 10K
- catalogue products updated daily
- 200K+
- video hours processed monthly
- 500K+
- daily analytics users
- ~5M
- platform uptime
- 99.9%
- activity events per day
- ~200K
Stage 1 · Olive Theory · Apr–Oct 2017
Camera-to-product discovery
The first version turned a user's photo into a shopping action. I owned the mobile architecture and hands-on React Native delivery, together with the catalogue data path needed to keep more than 200,000 products current each day.
Capture intent
A shopper takes a product photo in the React Native application.
Match products
The visual-matching service searches the current product catalogue for similar items.
Return choices
Ranked, purchasable products are presented back in the mobile shopping flow.
Stage 2 · Accio.ai · Nov 2017–Jul 2022
Video context to in-player commerce
Accio expanded the visual-discovery problem from a single user photo to large-scale video processing. I worked hands-on across platform architecture and delivery while leading the engineering effort that made contextual products available through the video player.
Analyse media
Video-processing services derive usable visual context while keeping the customer-facing player decoupled.
Use the pause moment
A viewer pause becomes an explicit intent event for retrieving contextually relevant products.
Deliver and observe
The player widget presents similar products while analytics track activity, reach and reliability.
Architecture decisions that carried forward
Reuse the matching foundation
The product-discovery capability began with a camera-driven mobile experience. Accio extended the same core problem—finding relevant products from visual context—into video rather than treating it as a disconnected rewrite.
Separate media, matching and delivery
Media capture, product matching, catalogue updates, analytics and the customer-facing surface were kept as distinct responsibilities. That made the system easier to evolve from a mobile search flow into an in-player experience.
Design around the user moment
Olive Theory started with an explicit camera action. Accio used the viewer's pause event as the intent signal, then returned visually relevant products inside the player without breaking the viewing flow.
Make scale observable
The platform added operational and product analytics as traffic grew. Reliability, video-processing volume and user activity were treated as product constraints, not retrospective reporting.
What I personally owned
- • React Native mobile architecture and delivery
- • Product-catalogue data pipelines
- • Visual-discovery platform evolution
- • Contextual video-commerce delivery
- • Distributed analytics architecture
- • Engineering leadership through scale
Employment timeline: Olive Theory ended in October 2017; Accio.ai began in November 2017. The continuity is the technology and product problem, not overlapping employment.