Scalable GPU-Powered Card Recognition Backend
Budget: $250 – $750 USD
I’m building an iOS app that recognises Pokémon cards from a photo, pulls live prices, and lets players curate their collections. To support it, I need a production-ready backend whose first priority is raw speed and horizontal scalability.
Core requirements
• GPU-accelerated image-embedding pipeline so card recognition stays sub-second at launch volumes and can scale without code changes.
• Strict, versioned REST contracts:
– POST /images → returns card ID + confidence
– GET /cards/{id}/price → returns current price from PokePriceTracker
– CRUD endpoints for user collections and profiles
• Secure, modern authentication (OAuth 2.0 or JWT are both fine) with refresh-token flow and role-based access.
• Stateless, modular architecture so the same APIs power iOS, future web, and desktop clients without duplication.
• Cloud-friendly infrastructure; containerised services that I can deploy to AWS, GCP or similar with autoscaling turned on.
Deliverables
1. Source code and Dockerfiles for all microservices.
2. Infrastructure-as-Code scripts to spin everything up (Terraform or CloudFormation).
3. OpenAPI 3.0 spec covering every endpoint.
4. Short performance report showing recognition latency and throughput under load.
5. Setup guide so my iOS team can hit the staging server immediately.
If you’ve built GPU-backed, high-traffic APIs before and are comfortable threading image models with real-time pricing data, let’s talk.
Core requirements
• GPU-accelerated image-embedding pipeline so card recognition stays sub-second at launch volumes and can scale without code changes.
• Strict, versioned REST contracts:
– POST /images → returns card ID + confidence
– GET /cards/{id}/price → returns current price from PokePriceTracker
– CRUD endpoints for user collections and profiles
• Secure, modern authentication (OAuth 2.0 or JWT are both fine) with refresh-token flow and role-based access.
• Stateless, modular architecture so the same APIs power iOS, future web, and desktop clients without duplication.
• Cloud-friendly infrastructure; containerised services that I can deploy to AWS, GCP or similar with autoscaling turned on.
Deliverables
1. Source code and Dockerfiles for all microservices.
2. Infrastructure-as-Code scripts to spin everything up (Terraform or CloudFormation).
3. OpenAPI 3.0 spec covering every endpoint.
4. Short performance report showing recognition latency and throughput under load.
5. Setup guide so my iOS team can hit the staging server immediately.
If you’ve built GPU-backed, high-traffic APIs before and are comfortable threading image models with real-time pricing data, let’s talk.