Design a Market Surveillance Platform
Company: Paxos
Role: Software Engineer
Category: System Design
Difficulty: medium
Interview Round: Onsite
Design a customer-facing market surveillance platform for financial exchanges.
Financial exchanges submit trading orders to the platform. An example order is: trader `T4` places an order to buy `1 BTC` at `30,000 USD`.
The platform must identify potential market manipulation activity. Detection is performed by pre-built algorithms that consume incoming orders and asynchronously generate suspected manipulation events with related metadata, such as orders, traders, exchanges, timestamps, and reason codes.
There are two detection algorithms:
1. **Trader-keyed algorithm**: every order for a given trader must be processed by the same algorithm instance.
2. **Exchange-keyed algorithm**: every order from a given exchange must be routed to a single algorithm instance responsible for that exchange.
Additional constraints:
- Trader identifiers are globally unique across all exchanges.
- Each algorithm can be independently replicated and deployed.
- Each algorithm instance may process multiple traders or exchanges, depending on throughput capacity.
- You do not need to design the internal detection algorithms.
- The system must handle high QPS.
- The design should consider availability, reliability, durability, and scalability.
Design the end-to-end system, including APIs, ingestion, routing, storage, asynchronous processing, failure handling, and how suspected manipulation events are exposed to customers.
Quick Answer: This question evaluates system design and distributed-systems competencies, including high-throughput ingestion, stateful routing guarantees, partitioning and replication strategies, asynchronous processing, fault tolerance, and data durability for detection pipelines.