MLOps Engineer

Job Opening: MLOps Engineer

Location: Bangalore

Employment Type: Full-time (Hybrid)

Role Overview

We are looking for an experienced MLOps Engineer to manage and optimize the complete machine learning lifecycle. The role involves supporting large-scale distributed training on on-premise GPU clusters and deploying optimized models into production-grade C++ inference systems.

Key Responsibilities:

High-Performance Inference & Deployment

• Build and maintain pipelines to move models from Python research environments to production C++ systems
• Optimize deep learning models (PyTorch/JAX/TensorFlow) for low-latency inference using quantization, operator fusion, TensorRT, or custom CUDA kernels
• Ensure deterministic behavior and thread safety in real-time environments

Research Infrastructure & Compute

• Manage and scale on-premise HPC/GPU clusters
• Optimize scheduling and resource utilization using Slurm or Kubernetes
• Troubleshoot distributed training issues (NCCL, InfiniBand, GPUDirect)

Data & Feature Engineering Pipelines

• Design high-throughput data loaders for large-scale historical datasets
• Implement feature stores ensuring consistency between offline training and online inference

Reliability & Monitoring

• Develop real-time monitoring, drift detection, and automated safety mechanisms
• Build CI/CD workflows for code, data, and model artifacts with strict reproducibility

Required Qualifications

• 3+ years of experience in MLOps, Systems Engineering, or SRE
• Strong proficiency in Python and working knowledge of C++
• Hands-on experience with Docker and Kubernetes or Slurm
• Strong understanding of Linux internals (memory, CPU pinning, NUMA, networking)
• Experience with PyTorch or TensorFlow and model serialization (ONNX, TorchScript)

Preferred Skills

• Experience managing large-scale on-premise GPU clusters (A100/H100)
• CUDA kernel optimization or experience with Triton/TVM
• Exposure to time