ML Ops Engineer — Agentic AI Lab (Founding Team)

Sourced from feeny.ai

Fabrion's Agentic AI Lab is hiring a founding MLOps Engineer to bridge ML research and production systems. You'll automate model training, deployment, versioning, and observability pipelines for open-source LLMs, RAG, and agent-native infrastructure. The role spans GPU compute orchestration, model governance, security, and evaluation frameworks. Backed by 8VC, this is a founding-tier position with meaningful equity.

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

• 4+ years in MLOps, ML platform engineering, or infra-focused ML roles
• Deep familiarity with model lifecycle tools: MLflow, Weights & Biases, DVC, HuggingFace Hub
• Experience deploying large open-source LLMs (LLaMA, Mistral, Falcon, Mixtral)
• Comfortable with tuning libraries: HuggingFace Trainer, DeepSpeed, FSDP, QLoRA
• Proficiency with Terraform, Helm, Kubernetes, and container orchestration
• Experience with CI/CD for ML and hybrid GPU cloud workloads
• Familiarity with LangChain, LangGraph, LlamaIndex, and vector databases
• Python as primary language; Bash required, Rust or Go a plus

• Experience with SOC2, HIPAA, or GovCloud-grade model operations
• Prior founding engineer or early-stage startup experience
• Experience with policy-as-code systems (OPA, Rego)
• Cost optimization for GPU workloads (spot instances, model sharding)