JD:

Those are the skill sets we need, hands-on with:

vLLM / LLM inference serving

GPT-OSS / Harmony parser troubleshooting

Tool/function calling debugging

LLM token, context-window and output parsing

Prompt caching / KV-cache optimization

NVIDIA GPU monitoring and performance tuning

Tensor Parallelism (TP2/TP4) tuning

Concurrency, batching and throughput optimization

Python-based AI application debugging

Docker / Kubernetes production operations

API Gateway / reverse proxy troubleshooting

HTTP troubleshooting

TLS / certificate / connection-reset debugging

Load balancer troubleshooting – ALB/NLB

OpenTelemetry / distributed tracing

Langfuse or equivalent LLM observability

Centralized logging / Kibana

End-to-end correlation IDs and request tracing

Retry, timeout, backoff and circuit-breaker design

Fallback / graceful-degradation patterns

Load, stress and soak testing

Latency analysis: P50/P95/P99, TTFT, tokens/sec

Capacity planning and saturation analysis

Production incident management and RCA

CI/CD, configuration versioning and rollback