T-Hub - AIOps Engineer - AI Infrastructure & Orchestration

Brak informacji o wynagrodzeniu

SeniorOn-site·WarszawaFull-time·Umowa o pracę

#480147·Dodano wczoraj·0

Źródło: justjoin.it

Aplikuj teraz

Tech Stack / Keywords

DevOpsSREMLOpsLLM GitLab Jenkins

Podsumowanie oferty

Design, deploy, and maintain LLM inference services on OpenShift/Kubernetes running on bare-metal GPU infrastructure. Manage model lifecycles, performance, monitoring, and secure operations for AI inference services.

Wymagania

5+ years of experience in DevOps, Site Reliability Engineering (SRE), Platform Engineering, or Infrastructure Operations.
At least 2 years of hands-on experience supporting MLOps, AI Infrastructure, or Large Language Model (LLM) platforms.
Strong experience administering Kubernetes and OpenShift in production environments.
Proven experience deploying and operating vLLM-based inference platforms in production.
Strong understanding of LLM serving concepts, including Paged Attention, continuous batching, and inference optimization techniques.
Deep knowledge of NVIDIA GPU technologies, CUDA drivers, NVIDIA Container Toolkit, and GPU troubleshooting.
Hands-on experience with Prometheus, Grafana, OpenTelemetry, and ELK Stack.
Experience building observability solutions, including custom metrics, exporters, dashboards, and alerting mechanisms.
Strong Python programming skills and experience developing automation and operational tooling.
Experience with Bash scripting and Linux systems administration.
Familiarity with GitLab CI, Jenkins, ArgoCD, and Infrastructure-as-Code practices.
Strong analytical and problem-solving skills in complex, distributed environments.
Excellent communication and collaboration skills.

Obowiązki

Design, deploy, and maintain vLLM inference services on OpenShift/Kubernetes running on bare-metal GPU infrastructure.
Manage NVIDIA GPU resources, including partitioning and allocation, to maximize utilization across models and tenants.
Automate model onboarding, versioning, deployment, hot-swapping, and rollback from private registries such as Hugging Face Enterprise and S3.
Implement and manage Horizontal Pod Autoscaling (HPA) based on workload demand, queue depth, and GPU utilization; optimize vLLM configurations and serving parameters.
Build and maintain observability and monitoring for AI inference services, including metrics collection, logging, and tracing.
Instrument vLLM endpoints to expose metrics for token consumption, latency, throughput, and error rates.
Develop usage tracking by API key, user, team, or department to support quota management and chargeback/showback requirements.
Create and maintain Grafana dashboards for infrastructure health, GPU utilization, inference performance, service availability, and business consumption metrics.
Configure proactive monitoring and alerting with Prometheus and Alertmanager to detect infrastructure failures, performance degradation, and unusual consumption patterns.
Implement and maintain API Gateway solutions for authentication, authorization, rate limiting, and intelligent routing to inference services.
Ensure secure operation of AI services through network segmentation, ingress and egress controls, and security best practices.
Maintain audit logging for compliance, security investigations, and operational governance.
Collaborate with AI Engineering, Platform Engineering, Security, and Infrastructure teams to deliver reliable, scalable, and secure AI services.
Participate in troubleshooting, incident response, root cause analysis, and continuous platform improvement.

Benefity

No dress code.
Medical, sport, and life insurance packages at preferential terms.
Access to the Percipio, Coursera, and Rodos learning platforms.
Access to company products and services at preferential terms.
Training or a financial bonus for recommending new employees.

Opieka zdrowotna

Karta sportowa

Ubezpieczenie

Premie

Inne informacje

The screening meeting lasts about 20–30 minutes.
The process may include one or two technical meetings with the Hiring Manager and/or team members, lasting 30–90 minutes each.

T-Mobile

95 aktywnych ofert

Zobacz wszystkie oferty

Aplikuj teraz