NIP-5: Scalable AI Job Pricing and Matching Modules

[kezeng2024]

Status: Approved

TL;DR

The project aims to design pricing and matching modules akin to stock exchange markets for AI developers and GPU miners. The pricing module includes bid and ask orders, while the matching module aims to match bids and asks at a predefined cadence. The design involves structuring orders, maintaining market prices, and managing order statuses.

The implementation includes implementing fundamental functionalities initially and later performance tuning.

Scope

The project entails creating pricing and matching modules for bid clients (AI developers) and ask clients (GPU miners). It involves defining bid and ask models, structuring orders, managing market prices, and implementing order matching functionalities.

The Design

1. Pricing Module: Utilizes bid and ask models, converting them into order structs. Market prices are maintained for GPU models. Methods include adding bids/asks, retrieving market prices, setting market prices, and canceling orders.

2. Matching Module: Matches bid and ask records periodically or another cadence. Final prices are determined by comparing bid and ask prices. Methods involve matching orders based on bid and ask records.

Timeline

• Stage 1 (Week of 05/20): Implement basic functionalities including placing bids/asks, updating market prices, and matching orders.

• Stage 2 (Week of 05/27): E2E test and performance tuning.

Future Work

Future enhancements could include optimizing matching algorithms, implementing order limit policies, and refining communication with clients regarding matched orders. Additionally, further scalability and performance improvements could be explored as the platform grows.