Design ZK Proving System - System Design | WinJob

Problem Statement: Industrial ZK Proving Platform

Design a production ZK proving platform for rollups and proof-as-a-service: witness ingestion, GPU proving clusters, recursive aggregation, and Ethereum verifier integration. This section covers problem statement in the understanding phase.

Operational detail

Anchor scale: 2–8k proofs/hour per GPU class, witness blobs 50–500 MB, L1 verify gas 250k–900k depending on recursion depth, p95 queue wait under 15 minutes for consumer withdrawals.

Failure and edge cases

Unique to sec-01: Toxic waste mishandling, witness truncation, verifier contract drift, and prover GPU fleet stranding during blob fee spikes.

Interview checkpoints

javaOne Dark Pro

1  
2 public record ProofJob(String jobId, byte[] witnessHash, String circuitId) {}
3

pythonOne Dark Pro

1  
2 def witness_size_mb(tx_count: int, avg_witness_kb: int) -> float:
3     return tx_count * avg_witness_kb / 1024
4

typescriptOne Dark Pro

1  
2 export interface ProofJob { id: string; circuitId: string; priority: 'withdrawal' | 'batch' | 'tenant'; }
3

Why interviewers care

ZK Proving System interviews reward crisp scope, explicit trade-offs, and failure stories—not generic microservice diagrams.

Interview checkpoint

Name one failure story for Problem Statement: Industrial ZK Proving Platform that proves you understand real outages, not happy-path diagrams.

Key Highlights

•Prove arbitrary rollup transitions with STARK or PLONK-family backends
•Separate witness generation from polynomial commitment work
•Verify succinct proofs on Ethereum with fixed gas budgets
•Serve internal rollups and external proof-as-a-service tenants

1
2	public record ProofJob(String jobId, byte[] witnessHash, String circuitId) {}
3

1
2	def witness_size_mb(tx_count: int, avg_witness_kb: int) -> float:
3	return tx_count * avg_witness_kb / 1024
4

1
2	export interface ProofJob { id: string; circuitId: string; priority: 'withdrawal' \| 'batch' \| 'tenant'; }
3