Senior Low-Latency Network Performance Engineer — Node.js / HTTP / TCP

We are looking for a senior low-latency / network performance engineer to analyze and optimize an existing Node.js HTTP automation system. The system is already working in production. We are NOT looking for a developer to build a new application from scratch. We need an expert who can measure the existing system at a very low level, identify the real latency bottlenecks, and optimize the critical request path. Current environment: - Node.js 24 on Windows/VPS - HTTP/1.1 - fetch() / Undici - Persistent HTTP connections / Keep-Alive - nginx-based remote service - VPS located very close to the remote service - Network latency is already very low The application continuously polls an authorized calendar/availability endpoint. When availability is detected, it immediately sends a prepared HTTP POST request. The POST URL, headers, authentication/session data and request body are prepared before availability is detected. We want as little work as possible between detection and the actual network send. Some measurements from the current system: Calendar GET responses can be approximately: 2–7 ms under good conditions. Detection → fetch() call: typically sub-millisecond to a few milliseconds, but occasionally higher. Booking POST measurements can look like: TRUE → FETCH CALL: 0.919 ms FETCH CALL → HEADERS: 117.509 ms HEADERS → BODY: 1.996 ms TOTAL: 120.423 ms In other tests, response-header latency can vary significantly. A dedicated Undici keep-alive test showed approximately: Cold request: HEADERS: 230+ ms Warm requests: HEADERS: approximately 10–25 ms A connection kept idle for 60 seconds was still able to perform a subsequent request at approximately 40 ms to response headers. We need someone who understands that response-header latency is NOT necessarily the same as request-send latency. We want proper instrumentation of the entire critical path. We need you to investigate: 1. Actual request-send timing at socket/network level. 2. TCP/TLS connection reuse and keep-alive behavior. 3. Node.js fetch() vs direct Undici Client/Pool overhead. 4. Event-loop latency and scheduling jitter. 5. DNS/TCP/TLS overhead where applicable. 6. HTTP redirects and their impact on our existing measurements. 7. Whether connections are genuinely reused for the critical POST. 8. Packet-level timing using Wireshark/tcpdump or equivalent tools. 9. p50/p95/p99 latency rather than relying on individual measurements. 10. Any unnecessary work between receiving availability and transmitting the POST. 11. Safe architectural/code improvements that reduce latency and jitter without increasing request volume or bypassing service protections. Ideal background: - Senior/Principal backend or network performance engineer - Low-latency systems - Node.js internals and Undici - HTTP/1.1 and HTTP/2 - TCP/IP and TLS - Linux networking - Wireshark/tcpdump - Performance profiling - Race-condition / high-concurrency systems - Experience with trading, real-time systems, gaming infrastructure, ad-tech, or other latency-sensitive systems is a strong plus. IMPORTANT: This is a performance engineering task on our authorized client workflow. We are not requesting unauthorized access, exploitation, bypassing security controls, or bypassing rate limits. We can provide the existing source code, timing logs, VPS environment and test cases. For your application, please answer: 1. How would you determine when the POST bytes actually leave the machine, rather than when fetch() is called or response headers arrive? 2. Which tools would you use to separate application latency, TCP/TLS latency, network latency and remote-server processing time? 3. Have you optimized systems where milliseconds matter? Please describe one example. 4. Are you available to begin immediately? We are willing to pay well for a genuinely senior specialist who can demonstrate the bottleneck with measurements and improve the system.