Qdrant • 📍 Europe / US Remote •
$175k - $235k USD
Design disk-backed HNSW graph traversal over Linux io_uring, tune Raft shard consensus rebalancing, and write hand-vectorized distance functions.
#raft • $176k-$239k USD Base
Raft in Production Rust: Architecture & Open Roles
Distributed consensus state replication for fault-tolerant storage clusters
1. Why Teams Deploy Raft
Every distributed database or coordination service needs linearizable consensus for metadata placement, leader election, and write-ahead log replication. In Rust, production systems rely on crates like `openraft` (used by Databend), PingCAP's `raft-rs` (extracted from TiKV), or custom Paxos/Raft state machines inside Neon, Qdrant, and Cockroach/Redpanda-style architectures.
2. Raft Kernel & Memory Internals
High-throughput storage engines never run a single monolithic Raft group for an entire node. Instead, they shard keys into thousands of independent ranges (`Multi-Raft`), coalescing heartbeats and batching log appends to amortize `fsync` latency. Engineers test consensus implementations using deterministic simulation testing (`madsim` or `turmoil`) to inject network partitions and clock drift.
3. Raft Interview & Hiring Bar
Distributed database vendors pay $180k to $260k for engineers who understand Raft log compaction, snapshot streaming, read-index optimizations (avoiding disk writes on linearizable reads), and split-brain prevention.
Open Raft Engineering Positions (4)
Neon (Databricks) • 📍 Worldwide Remote •
$190k - $265k USD
Improve L0-to-L1 compaction throughput over `tokio-epoll-uring`, lower tail latency on cold page reconstructions, and harden Paxos WAL quorum recovery.
Turso • 📍 Worldwide Remote •
$175k - $235k USD
Implement SQLite bytecode VDBE instructions in safe Rust, expand deterministic simulation fuzzing seeds, and reduce page-cache lock contention.
