About the role
About Us
Parametric builds autonomous physical businesses. We're starting with a laundromat, designed from the ground up for autonomy. Dirty laundry in, folded clean laundry out. Rather than retrofit human workflows with robots, we redesign processes with robots as first class citizens.
About the Role
As a Senior ML Infrastructure Engineer, you will build the distributed nervous system of our research operations. You will design the underlying architecture that transforms raw robot logs into deployed policies, managing the complexity of asynchronous workflows at scale.
This is a deep systems engineering role. You will be responsible for orchestrating the interplay between massive data ingestion, large-scale distributed training, and parallelized simulation. You will build the tooling that ensures our experiments are reproducible, our compute is saturated, and our lineage is traceable.
What You’ll Do
- Orchestrate Complex DAG Workflows: Design and implement fault-tolerant, multi-stage DAGs (using Airflow, Argo, or Temporal) that coordinate asynchronous tasks across distributed systems—from log ingestion and synthetic data generation to training and evaluation.
- Architect Parallelized Compute Clusters: Provision and manage high-performance compute clusters (Kubernetes/Slurm) tailored for multi-node distributed training. You will optimize for scheduling efficiency, handling spot instance interruptions and maximizing GPU interconnect throughput.
- Engineer A/B Experimentation Infrastructure: Build the platform for rigorous model comparison, enabling simultaneous A/B testing of policies in simulation and real-world shadow mode. You will implement automated statistical profiling to gate deployments based on key performance metrics.
