Palo Alto, CA | Full-Time | On-site
Role Overview:As a Senior MLOps Engineer, you will own the infrastructure that takes Nace.AI's models from research to reliable, production-grade systems. Our infrastructure generates task-specific Small Language Models (SLMs) in real time - which means our training, serving, and evaluation infrastructure isn't an afterthought; it is the product. You will design and operate the pipelines, orchestration, and serving layers that allow us to train, deploy, monitor, and continuously improve many specialized models at once, with the reliability that high-stakes audit, compliance, and finance workflows demand. This role sits at the intersection of ML engineering, LLM inference infrastructure, and platform reliability, and requires both strong systems instincts and hands-on execution.
Key Responsibilities:- Design, build, and operate end-to-end ML infrastructure: training orchestration, experiment tracking, model registries, CI/CD for models, and automated evaluation pipelines.
- Own LLM/SLM serving infrastructure - scale low-latency, high-throughput inference using frameworks like vLLM, including batching, caching, and autoscaling strategies.
- Build and manage multi-GPU training and inference clusters (scheduling, utilization, cost optimization) across cloud and on-prem environments.
- Implement observability for models in production: latency, throughput, drift, regression, and quality monitoring with actionable alerting.
- Apply inference-time optimizations - quantization (AWQ, GPTQ, FP8/GGUF), distillation support, KV-cache management, and deployment tuning - in partnership with our ML and Research Engineers.
- Harden our stack for enterprise deployment: reproducibility, versioning, access controls, and audit-ready traceability of model behavior.
- Set MLOps best practices and tooling standards as an early, senior member of the infrastructure team.
Qualifications:- 5+ years of experience in MLOps, ML infrastructure, or platform engineering, with substantial production ownership.
- Proven experience deploying and scaling LLM, inference infrastructure in production, including model serving frameworks such as TRT, vLLM, SGLang or TGI.
- Strong proficiency with Kubernetes, containerization (Docker), and infrastructure-as-code (Terraform or similar).
- Hands-on experience with GPU cluster management and distributed training/serving environments.
- Proficient in Python with a strong track record of building substantial, maintainable systems.
- Experience with ML pipeline and orchestration tooling (e.g., Airflow, Kubeflow, Ray, MLflow, Weights & Biases).
- Solid foundation in computer science fundamentals and cloud architecture (AWS, GCP, or Azure).
- BS degree in CS or related technical field.
- Self-starter comfortable working in a fast-paced, dynamic environment.
Preferred Qualifications:- MS in CS or related technical field.
- Experience operating multi-node GPU training infrastructure.
- Hands-on experience with quantization techniques (AWQ, GPTQ, FP8/GGUF) and other inference-time optimizations.
- Familiarity with data processing stacks such as Spark and Airflow.
- Experience supporting fine-tuning workflows for LLMs/VLMs (instruction tuning, RLHF/DPO pipelines).
- Experience in regulated or enterprise environments where reliability, security, and auditability are first-class requirements.
- Contributor to open-source ML infrastructure projects.
