Machine Learning Engineer (IC4/IC5)
About the Role
You will own the end-to-end ML model lifecycle from post-training through production — everything after the researchers hand off a trained model. This is not a research role. You are the engineer who takes models and makes them real: benchmarked, deployed, monitored, and integrated into live production applications. You will work directly with ML researchers, production engineers, and platform teams in a fast-moving hybrid cloud environment.
What You Will Do
Inference & Deployment
Evaluate and benchmark new ML inference frameworks to guide production decisions
Deploy models to Google Cloud Platform and integrate them into production applications and Java-based streaming pipelines
Own deployment automation end-to-end — from model handoff through live serving
Monitor how models behave in production for real end-users
Performance & Quality
Design and execute benchmarking, performance testing, and quality testing on ML models
Perform model sampling to support quality evaluation and researcher feedback loops
Debug issues across the full stack — from inference layer down to streaming pipelines
Cross-functional Collaboration
Partner with ML researchers to provide benchmarking feedback and guide inference decisions — requires enough core ML knowledge to have a meaningful technical handshake
Adapt rapidly to non-standard and evolving tech stacks across hybrid (on-prem + Google Cloud Platform) infrastructure
Technical Stack
Primary platform: Google Cloud Platform (inference, deployment automation, experimentation, sampling)
Production integration: Java-based streaming pipelines (model integration layer)
Infrastructure: Hybrid — on-premise streaming + Google Cloud Platform serving stacks
Distributed systems: Working knowledge required for debugging and end-to-end testing (not deep expertise)
Machine Learning frameworks (TensorFlow, PyTorch, JAX or similar)
What We Are Looking For
Must-Have
Strong foundation in ML inference, deployment, and quality testing
Demonstrated ability to ramp up quickly on new and unfamiliar tech stacks — this is the single most important trait
End-to-end problem-solving mindset — can own a problem from model handoff to user-facing behavior
Core ML knowledge sufficient to benchmark models and collaborate with researchers
Experience deploying models in cloud environments, ideally Google Cloud Platform
Good to Have
Exposure to Java or JVM-based systems (model integration happens in Java; deep expertise not required)
Familiarity with streaming data architectures
Experience in hybrid cloud/on-prem environments