ML Ops Engineer — Agentic AI Lab (Founding Team)

About the role

Our AI Lab is pioneering the future of intelligent infrastructure through open-source LLMs, agent-native pipelines, retrieval-augmented generation (RAG), and knowledge-graph-grounded models. We’re hiring an ML Ops Engineer to be the glue between ML research and production systems — responsible for automating the model training, deployment, versioning, and observability pipelines that power our agents and AI data fabric.

Responsibilities

• Build and maintain secure, scalable, and automated pipelines for: LLM fine-tuning, SFT, LoRA, RLHF, DPO training, RAG embedding pipelines with dynamic updates, model conversion, quantization, and inference rollout
• Manage hybrid compute infrastructure (cloud, on-prem, GPU clusters) for training and inference workloads using Kubernetes, Ray, and Terraform
• Containerize models and agents using Docker, with reproducible builds and CI/CD via GitHub Actions or ArgoCD
• Create and manage evaluation and benchmarking frameworks (e.g. OpenLLM-Evals, RAGAS, LangSmith)
• Implement and enforce model governance: versioning, metadata, lineage, reproducibility, and evaluation capture
• Integrate with security and access control layers (OPA, ABAC, Keycloak) to enforce model policies per tenant
• Instrument observability for model latency, token usage, performance metrics, error tracing, and drift detection
• Support deployment of agentic apps with LangGraph, LangChain, and custom inference backends (e.g. vLLM, TGI, Triton)

Requirements

• Model Infrastructure: 4+ years in MLOps, ML platform engineering, or infra-focused ML roles
• Deep familiarity with model lifecycle management tools: MLflow, Weights & Biases, DVC, HuggingFace Hub
• Experience with large model deployments (open-source LLMs preferred): LLaMA, Mistral, Falcon, Mixtral
• Familiarity with tuning libraries (HuggingFace Trainer, DeepSpeed, FSDP, QLoRA)
• Familiarity with inference serving: vLLM, TGI, Ray Serve, Triton Inference Server
• Automation + Infra: Proficient with Terraform, Helm, K8s, and container orchestration
• Experience with CI/CD for ML (e.g. GitHub Actions + model checkpoints)
• Managed hybrid workloads across GPU cloud (Lambda, Modal, HuggingFace Inference, Sagemaker)
• Familiar with cost optimization (spot instance scaling, batch prioritization, model sharding)
• Agent + Data Pipeline Support: Familiarity with LangChain, LangGraph, LlamaIndex or similar RAG/agent orchestration tools
• Built embedding pipelines for multi-source documents (PDF, JSON, CSV, HTML) integrated with vector databases (Weaviate, Qdrant, FAISS, Chroma)
• Security & Governance: Implemented model-level RBAC, usage tracking, audit trails integrated with API rate limits, tenant billing, and SLA observability
• Experience with policy-as-code systems (OPA, Rego) and access layers

Desired Experience

• Preferred Stack: HuggingFace, DeepSpeed, MLflow, Weights & Biases, DVC, Infra: Kubernetes (GKE/EKS), Ray, Terraform, Helm, GitHub Actions, ArgoCD, Serving: vLLM, TGI, Triton, Ray Serve, Pipelines: Prefect, Airflow, Dagster, Monitoring: Prometheus, Grafana, OpenTelemetry, LangSmith, Security: OPA (Rego), Keycloak, Vault, Languages: Python (primary), Bash, optionally Rust or Go for tooling

Mindset & Culture Fit

• Builder's mindset with startup autonomy: you automate what slows you down
• Obsessive about reproducibility, observability, and traceability
• Comfortable with a hybrid team of AI researchers, DevOps, and backend engineers
• Interested in aligning ML systems to product delivery, not just papers

Why This Role Matters

Your work will enable models and agents to be trained, evaluated, deployed, and governed at scale — across many tenants, models, and tasks. This is the backbone of a secure, reliable, and scalable AI-native enterprise system. If you dream about using AI to solve some really hard real world problems – we would love to hear from you.