Async Production Systems

required2,475 repos of evidence

Infrastructure roles assume backend fluency — the corpus's infra projects are async services first, distributed systems second.

The destination — non-negotiable
  • operate (not just build) long-running async services
  • write code that assumes the network will fail
  • debug a task leak or a stuck future
+ beyond Cargo: everything from the Backend path's first two legs · deadlines everywhere · graceful degradation
The vehicles — 2 ways in, any combination works

Long-running daemons operated by real organizations — read their shutdown, retry, and failure paths.

Prove it — build this

If you haven't: complete the Backend path's API project — it's the substrate everything below runs on.

Can you honestly tick these?

Networking Under Failure

required1,122 repos of evidence

Every distributed failure is a network failure first. The corpus's proxies, DNS servers, and transports are where those failures live in readable code.

The destination — non-negotiable
  • understand network failures: partitions, timeouts, retries, backpressure
  • understand what proxies and load balancers actually do to your traffic
  • reason about latency budgets across hops
+ beyond Cargo: TCP vs QUIC tradeoffs · connection pooling & reuse · retry storms & jitter · health checking
The vehicles — 3 ways in, any combination works

Production proxies and mesh dataplanes — the code your packets already flow through.

monosans/proxy-scraper-checkerapproachable★ 1.3k · pushed this monthHTTP, SOCKS4, SOCKS5 proxies scraper and checker with rich functionality.raine/claude-code-proxyapproachable★ 610 · 46 open issues · landableUse Claude Code with your ChatGPT or Kimi subscription via a local Anthropic-compatible proxyvalnesfjord/tg-ws-proxy-rsapproachable★ 188 · pushed this weekTelegram MTProto WebSocket Bridge Proxy — a Rust vibecoded port of Flowseal/tg-ws-proxylinkerd/linkerd2-proxyintermediate★ 2.1k · run by Buoyant · 20 open issues · landableA purpose-built proxy for the Linkerd service mesh. Written in Rust.xin2017338/lynx-proxyintermediate★ 502 · pushed this monthLynx Proxy is a high-performance and flexible proxy tool developed in RustEmbarkStudios/quilkinintermediate★ 1.6k · run by Embark Studios · 75 open issues · landableQuilkin is a non-transparent UDP proxy specifically designed for use with large scale multiplayer dedicated game server deployments, to ensure security, access control, telemetry data, metrics and more.rtk-ai/rtkadvanced★ 81k · pushed this month · curator pickCLI proxy that reduces LLM token consumption by 60-90% on common dev commands. Single Rust binary, zero dependenciesagentgateway/agentgatewayadvanced★ 4.9k · 273 open issues · landableNext Generation Agentic Proxy for AI Agents and MCP serversDevolutions/devolutions-gatewayadvanced★ 161 · pushed this weekA blazing fast relay server adaptable to different protocols and desired levels of traffic inspection.
Prove it — build this

Build a small L7 reverse proxy: health checks, per-backend circuit breaking, and a chaos script that kills backends while a load test runs.

Can you honestly tick these?

Coordination & Consensus

required65 repos of evidence

The corpus puts consensus in code you can read — implementations, coordination clients, and the databases built on them. Theory is table stakes; these repos are the source.

The destination — non-negotiable
  • understand consensus: what Raft guarantees and what breaks it
  • understand leader election, membership, and split brain
  • understand why exactly-once delivery is a lie and what to do instead
+ beyond Cargo: Raft mental model · quorums & fencing · clocks & ordering · idempotency as a design tool
The vehicles — 3 ways in, any combination works

Raft and consensus libraries themselves — read the state machine, not the blog post about it.

openraft21 reposraft8 repos
Prove it — build this

Build a replicated key-value store on a Raft crate: three nodes, leader failover test, and a README explaining exactly what is and isn't guaranteed.

Can you honestly tick these?

Queues & Streaming

required179 repos of evidence

Queues are how distributed systems absorb failure. The corpus's stream processors, log shippers, and messaging clients show the semantics in running code.

The destination — non-negotiable
  • understand queues: delivery guarantees, visibility timeouts, dead letters
  • understand streams versus queues and when each fits
  • design consumers that survive redelivery and reordering
+ beyond Cargo: at-least-once semantics · consumer groups & offsets · backpressure end-to-end · windowing basics
The vehicles — 2 ways in, any combination works

The engines that transform data in motion — windowing, watermarks, and exactly-once claims to audit.

parseablehq/parseableapproachable★ 2.5k · 19 open issues · landableParseable is an observability datalake built from first principles.volga-project/volgaapproachable★ 161 · pushed this monthReal-time data processing/feature engineering tailored for modern AI/ML systems.kaushiksrini/parqeyeapproachable★ 678 · 15 open issues · landablePeek inside Parquet files right from your terminalPRQL/prqlintermediate★ 11k · 256 open issues · landablePRQL is a modern language for transforming data — a simple, powerful, pipelined SQL replacementquickwit-oss/quickwitintermediate★ 12k · run by Quickwit · pushed this monthCloud-native search engine for observability. An open-source alternative to Datadog, Elasticsearch, Loki, and Tempo.cube-js/cubeintermediate★ 21k · pushed this month📊 Cube Core is open-source semantic layer for AI, BI and embedded analyticsvectordotdev/vectoradvanced★ 23k · pushed this month · curator pickA high-performance observability data pipeline.pola-rs/polarsadvanced★ 40k · run by Polars · pushed this monthExtremely fast Query Engine for DataFrames, written in Rustjuspay/hyperswitchadvanced★ 44k · pushed this month · curator pickOpen source, composable payments platform | PCI compliant | SaaS and Self-host options | Enables connectivity to multiple payment, payout, fraud, vault and tokenization providers | Uplifts authorization with intelligent routing and revenue recovery | Reduce payment processing costs with cost observability | Reduces payment ops with reconciliation
Prove it — build this

Build a work-queue system with at-least-once delivery, visibility timeouts, and a chaos test that kills workers mid-job.

Can you honestly tick these?

Storage Tradeoffs

recommended443 repos of evidence

Rust's storage-engine corpus is unusually strong — reading these codebases is the fastest route to credible storage conversations.

The destination — non-negotiable
  • understand storage tradeoffs: LSM vs B-tree, WALs, compaction
  • know what fsync guarantees and what durability actually costs
  • work against object storage semantics without pretending it's a filesystem
+ beyond Cargo: LSM trees vs B-trees · WALs & durability · compaction · object storage semantics
The vehicles — 3 ways in, any combination works

B-tree and LSM engines you can read end to end — the WAL, the compactor, the recovery path.

Prove it — build this

Write a tiny persistent KV store: WAL, crash-recovery test, and a README explaining your durability guarantees honestly.

Can you honestly tick these?

Observability & Reliability

required472 repos of evidence

Infra engineers are judged on behavior under failure. The corpus's reliability layers — tracing, otel, governor — are the tools of that judgment.

The destination — non-negotiable
  • propagate trace context across service boundaries
  • understand SLOs, error budgets, and load shedding
  • run async systems in production and learn from their incidents
+ beyond Cargo: SLOs & error budgets · distributed tracing · rate limiting & load shedding · incident habits
The vehicles — 3 ways in, any combination works

OpenTelemetry-instrumented systems and the collectors themselves — context propagation done for real.

parseablehq/parseableapproachable★ 2.5k · 19 open issues · landableParseable is an observability datalake built from first principles.apache/skywalking-query-protocolapproachable★ 72 · pushed this monthQuery Protocol for Apache SkyWalking in GraphQL formatDataDog/pupapproachable★ 1.0k · run by Datadog · 14 open issues · landableGive your AI agent a Pup — a CLI companion with 200+ commands across 33+ Datadog products.zama-ai/fhevmintermediate★ 25k · pushed this month · curator pickFHEVM, a full-stack framework for integrating Fully Homomorphic Encryption (FHE) with blockchain applicationsquickwit-oss/quickwitintermediate★ 12k · run by Quickwit · pushed this monthCloud-native search engine for observability. An open-source alternative to Datadog, Elasticsearch, Loki, and Tempo.DataDog/libdatadogintermediate★ 69 · run by Datadog · pushed this monthDatadog shared rust-based library. For now only used in other products (e.g. Ruby or PHP libraries).juspay/hyperswitchadvanced★ 44k · pushed this month · curator pickOpen source, composable payments platform | PCI compliant | SaaS and Self-host options | Enables connectivity to multiple payment, payout, fraud, vault and tokenization providers | Uplifts authorization with intelligent routing and revenue recovery | Reduce payment processing costs with cost observability | Reduces payment ops with reconciliationopenai/codexadvanced★ 125k · run by OpenAI · pushed this monthLightweight coding agent that runs in your terminaldenoland/denoadvanced★ 108k · run by Deno · pushed this monthA modern runtime for JavaScript and TypeScript.
Prove it — build this

Add distributed tracing across two services with OpenTelemetry — include the cross-service trace screenshot and an SLO doc.

Can you honestly tick these?
At the end of this road

Open roles matching this destination right now:

Radar's data platform stack uses TypeScript, Rust, and Python. Role focuses on building and scaling the data infrastructure powering 1B+ API calls per day.

Open
← All destinationsBrowse the evidence corpus →