Senior Machine Learning Infrastructure Engineer

Schrödinger•New York, NY

4d•$110,000 - $160,000

About The Position

Schrödinger is seeking a Senior Machine Learning Infrastructure Engineer to help transform the discovery of therapeutics and materials. As a member of the Machine Learning team, this role builds and maintains systems that put machine-learning models and AI agents to work in drug discovery. LiveDesign ML trains and serves models on live project data in Schrödinger’s clusters and customers’ LiveDesign environments. Bunsen is an agentic platform where each session runs an AI agent in its own isolated Kubernetes pod, and the company also operates locked-down environments for federated-learning collaborations with pharmaceutical partners. The engineer will own infrastructure across internal clusters and customer clouds on GCP and AWS. The focus is on repeatable customer deployments, strong isolation for AI-generated code, and reliable, cost-efficient GPU systems. The role works closely with machine-learning engineers and scientists.

Requirements

At least six years of infrastructure, platform, or SRE experience.
In-depth understanding of Kubernetes operation, beyond deployment.
Robust background with GCP and some familiarity with AWS.
Ability to write Terraform or OpenTofu and Helm and ship using GitOps.
Proficiency with Python and shell scripting.
Security-minded approach, including least privilege, managed secrets, and segmented networks.
Clear communication and ability to work closely with machine-learning engineers and scientists.

Nice To Haves

Experience running GPU workloads and model serving, such as vLLM and Knative, on Kubernetes.
Experience sandboxing untrusted code using technologies such as gVisor, Kata Containers, or Cilium.
Experience shipping software into customer-owned or restricted-egress clouds.
Experience with Argo Workflows or batch/HPC scheduling.
Experience in regulated environments, such as SOC 2, ISO 27001, or GxP.
Experience operating LLM or agent platforms.

Responsibilities

Turn customer installations and upgrades into versioned, GitOps-driven releases using Helm, OpenTofu, and Argo CD, with automated verification.
Design isolation for multi-tenant agent sandboxes, including network policy, workload identity, runtime sandboxing, and egress control.
Secure CI/CD and the software supply chain using keyless authentication, ephemeral runners, and pinned, signed images.
Build and operate locked-down partner environments with private networking, default-deny egress, just-in-time access, and audit logging.
Keep GPU-heavy training and inference reliable and cost-efficient, supported by observability, SLOs, and on-call.