Senior MLOps Engineer

• Creditas is looking for a Senior MLOps Engineer (2-5 yrs) to own the end-to-end infrastructure that powers our automatic speech recognition (ASR), next-generation speech-synthesis and voice agent products in Gurugram Office of Creditas

Key Responsibilities

• Design, deploy, and maintain cloud-native training & inference pipelines for large-scale ASR models
• Build high-throughput CI/CD workflows that automate data ingestion, model retraining, and A/B rollout
• Optimize distributed training on cloud GPU clusters, tuning mixed-precision, sharding, and checkpointing for maximum throughput
• Implement robust monitoring, alerting, and drift-detection for models in production
• Champion infrastructure as code (Terraform/CDK) and container orchestration (Kubernetes, Kubeflow, KubeGPT) for GPU workloads
• Collaborate with research teams to productionize speech-synthesis models and accelerate real-time inference

Minimum Qualifications

• BTech/MTech in CS, EE, or related field from IITs, NITs, IIITs, BITS, or other tier-1 institutes
• 2-4 years of hands-on MLOps experience supporting deep-learning workloads at scale
• Proficiency with Python, Docker, Kubernetes orchestration for GPU clusters, and at least one major cloud (AWS/GCP/Azure)
• Proven track record training or fine-tuning ASR models (eg, Conformer, Whisper, wav2vec 2.0)
• Deep familiarity with PyTorch Lightning or similar frameworks, plus experimentation trackers (MLflow, Weights & Biases)

Preferred Extras

• Prior work on speech-synthesis or TTS pipelines; knowledge of vocoders and diffusion-based models
• Experience optimizing inference for low-latency serving on cloud infrastructure
• Exposure to DataOps practices (feature stores, data quality testing)
• Security-first mindset with SOC 2 / ISO 27001 controls

What You'll Get

• Direct ownership of the MLOps backbone for products used by millions of end-users
• Access to cloud GPU infrastructure and compute budgets
• A fast-moving team obsessed with natural-sounding voices and lightning-fast inference
• Competitive compensation, flexible hybrid policy in our Gurgaon office, and an environment that values constant learning