Dubai Onsite Agentic AI Engineer: Ultra-Low Latency LLMs

Systems Ltd; Dubai; On-site; AED 360,000 - 600,000; Full time

Job summary

Systems Ltd in Dubai, UAE, is seeking an experienced Agentic AI Engineer to own end-to-end performance and reliability of enterprise LLM serving and agentic pipelines. You will manage vLLM inference, tune GPU clusters, and optimize caching and tensor parallelism across NVIDIA GPUs, with Docker and Kubernetes in production.

You will lead load, latency, and capacity tests, implement observability with OpenTelemetry, Langfuse, and Kibana, and collaborate with AI, DevOps, and SRE teams to ensure

Qualifications

• Bachelor's or Master's degree in Computer Science, Artificial Intelligence, Software Engineering, or related technical discipline.
• Extensive hands-on professional experience in AI infrastructure engineering, LLM inference serving, or Site Reliability Engineering for AI systems.
• Deep technical expertise in vLLM, LLM inference tuning, KV-cache optimization, and prompt caching mechanisms.
• Proven proficiency in NVIDIA GPU monitoring, performance tuning, and Tensor Parallelism configuration (TP2/TP4).
• Strong practical background in Python-based AI application debugging, concurrency management, and throughput optimization.
• Extensive production operations experience with Docker, Kubernetes, API gateways, and reverse proxy troubleshooting.
• Solid command of networking fundamentals including HTTP, TLS/SSL certificates, connection resets, and ALB/NLB load balancer administration.
• Practical experience implementing distributed tracing, OpenTelemetry, Langfuse observability, and Kibana centralized logging.
• Demonstrated ability in conducting load testing, capacity planning, saturation analysis, and high-stakes production incident management (RCA).
• Professional availability to work onsite in Dubai, UAE, with excellent communication and cross-functional leadership skills.

Responsibilities

• Deploy, configure, and manage vLLM and high-performance LLM inference engines in production environments.
• Troubleshoot and optimize GPT-OSS models and Harmony parsers for complex agentic tool and function-calling workflows.
• Implement prompt caching, KV-cache optimization, and advanced Tensor Parallelism (TP2/TP4) tuning strategies.
• Monitor NVIDIA GPU performance, memory allocation, and utilization metrics to maximize compute efficiency.
• Tune concurrency, dynamic request batching, and token generation throughput across distributed LLM clusters.
• Operate containerized AI applications and microservices using Docker and Kubernetes in enterprise production settings.
• Troubleshoot complex networking issues across API gateways, reverse proxies, HTTP protocols, TLS/certificates, and connection resets.
• Administer and debug load balancers (AWS ALB/NLB) and implement robust retry, timeout, backoff, and circuit-breaker patterns.
• Establish comprehensive observability using OpenTelemetry, distributed tracing, Langfuse LLM monitoring, and Kibana centralized logging.
• Conduct rigorous load, stress, and soak testing alongside granular latency analysis (P50/P95/P99, TTFT, tokens/sec) and capacity planning.

Skills

Python

vLLM

NVIDIA GPU tuning

Kubernetes

Docker

OpenTelemetry

Langfuse

Kibana

APIs / API gateways

Distributed tracing

Education

Bachelor's or Master's degree in CS/AI/SE

Tools

Docker

Kubernetes

API gateways

Reverse proxies

AWS ALB/NLB

Job description

Systems Ltd in Dubai, UAE, is seeking an experienced Agentic AI Engineer to own end-to-end performance and reliability of enterprise LLM serving and agentic pipelines. You will manage vLLM inference, tune GPU clusters, and optimize caching and tensor parallelism across NVIDIA GPUs, with Docker and Kubernetes in production.

You will lead load, latency, and capacity tests, implement observability with OpenTelemetry, Langfuse, and Kibana, and collaborate with AI, DevOps, and SRE teams to ensure