Senior ML Platform Reliability & Infrastructure Engineer

Holisticon Connect is a division within NEXER GROUP, a custom software development company started in Poland in 2017 with over 140 people. The team works on a highly advanced drug discovery platform at the intersection of machine learning, large-scale data systems, and computational science, building core infrastructure for AI-driven drug design to support new medicine discovery.

• 5+ years experience designing, operating, or scaling distributed multi-service architectures in production.
• Deep, hands-on experience with Kubernetes, especially GKE or equivalent managed Kubernetes.
• Comfortable with networking, RBAC, resource management, custom controllers, and debugging pod-level issues (OOMKilled, CrashLoopBackOff, scheduling failures).
• Proven track record building monitoring stacks using Prometheus, Grafana, Loki, and OpenTelemetry.
• Strong proficiency in Python, including FastAPI services, async GraphQL APIs, ML serving runtimes, and data pipelines.
• Experience with Infrastructure as Code using Terraform managing cloud resources at scale on GCP or AWS.
• Working knowledge of GCP services: GKE, Cloud SQL, Secret Manager, IAM.

Nice to have:

• Exposure to ML infrastructure including model serving frameworks (MLServer/Seldon, Ray/Anyscale), training pipelines, and model registry patterns (W&B or similar).
• Experience with message-oriented architectures such as Dapr, Redis Streams, or comparable pub/sub and event-driven patterns.
• Hands-on experience with workflow orchestration tools like Argo Workflows, Prefect, Airflow, or similar DAG-based pipeline engines.
• Familiarity with GraphQL APIs using Apollo Server or Strawberry GraphQL in production.
• Comfortable leading incident response and debugging sessions under time pressure.

• Profile and optimise inference latency and throughput for model-serving runtimes handling high-volume prediction requests behind a routing/gateway layer.
• Design and implement comprehensive observability across the platform by adding distributed tracing, effective logging, Grafana dashboards, alerting policies, and SLO/SLI frameworks using Prometheus, Loki, and OpenTelemetry.
• Harden Kubernetes workloads running on GKE by optimising GPU/CPU resource tuning and improving scaling of resources.
• Improve the resilience of asynchronous job pipelines built on Argo Workflows, Dapr pub/sub, and Redis, including retry strategies, dead-letter handling, and backpressure mechanisms.
• Collaborate with ML engineers and scientists to reduce friction in the model lifecycle from training and registration through to production serving.
• Lead debugging sessions across distributed components, correlating logs, traces, and metrics to identify root cause under time pressure.