Machine Learning Engineer

Role Overview

We are hiring a Machine Learning Engineer who will co-own the design, implementation, and operation of a fully on-premises ML platform built from scratch. In addition to developing models, you will be responsible for MLOps, infrastructure-aware ML pipelines, and production deployment on Kubernetes. This role goes beyond modeling you will help build the platform itself.

Key Responsibilities

• Design and implement end-to-end ML pipelines (data processing, training, evaluation, deployment) using Kubernetes Jobs.
• Build containerized ML workloads using Docker and define runtime environments.
• Deploy and maintain model serving services on Kubernetes (real-time and batch inference).
• Implement experiment tracking, model versioning, and reproducibility from scratch.
• Store and manage datasets, features, and model artifacts in MinIO.
• Integrate ML workflows with PostgreSQL for metadata, pipeline state, and model registry.
• Own MLOps workflows: training automation, model promotion, rollback, and re-training.
• Instrument ML workloads with Datadog for metrics, logs, traces, and alerts.
• Configure and secure ML workloads using Kubernetes ConfigMaps, Secrets, and RBAC.
• Optimize pipelines for performance, cost, and GPU usage.
• Collaborate closely with the Backend Engineer to define APIs, schemas, and system boundaries.
• Participate in infrastructure decisions, CI/CD setup, and production operations.

Required Qualifications

• Strong experience building and deploying production ML systems.
• Advanced proficiency in Python and ML frameworks (PyTorch, TensorFlow, scikit-learn).
• Hands-on experience with Docker and Kubernetes (not just using them, but configuring workloads).
• Experience with on-prem ML platforms.
• Familiarity with MinIO, PostgreSQL, Redis.
• Experience running ML workloads on GPUs.
• Exposure to CI/CD tooling and GitOps-style workflows.
• Prior experience as an MLOps or hybrid ML/Infra engineer.
• Experience building ML pipelines without managed cloud services.
• Understanding of experiment tracking, model lifecycle management, and reproducibility.
• Comfortable owning systems in production (debugging, monitoring, on-call mindset).