IICTI Certified MLOps Engineer

Holders demonstrate that they can apply and evaluate mL platform architecture and MLOps maturity strategy, reproducible pipelines, lineage and experiment tracking, feature stores and training-serving consistency and model registry governance and release promotion, plus 8 further areas, in realistic professional scenarios, as defined in the AQSCERT scheme for MLOps Engineer.

Who it is for: Professionals working, or preparing to work, in artificial intelligence roles.

1. 01ML platform architecture and MLOps maturity strategy4% of the exam

   • Architect ML platforms whose capabilities match organisational maturity, scale and risk profile.
   • Scope: Reference architectures for training, serving and feature layers, build vs buy (Vertex AI, SageMaker, Databricks, Kubeflow), MLOps maturity models, multi-tenancy and platform team operating models.
2. 02Reproducible pipelines, lineage and experiment tracking11% of the exam

   • Design training pipelines that are reproducible, auditable and traceable from data to deployed model.
   • Scope: Pipeline orchestration (Kubeflow Pipelines, Airflow, Dagster, Metaflow), data and model versioning (DVC, lakeFS), experiment tracking (MLflow, W&B), deterministic environments and end-to-end lineage metadata.
3. 03Feature stores and training-serving consistency9% of the exam

   • Design feature pipelines that guarantee consistency between training and serving and prevent leakage.
   • Scope: Online/offline feature stores (Feast, Tecton), point-in-time correct joins and leakage prevention, streaming feature computation, training-serving skew detection, feature reuse and ownership.
4. 04Model registry governance and release promotion8% of the exam

   • Lead governed model release processes with automated quality gates and safe progressive rollout.
   • Scope: Model registries and model cards, stage-based promotion and approval gates, signed artefacts and provenance (SLSA, Sigstore), automated evaluation gates, shadow, canary and champion-challenger rollout.
5. 05High-performance model serving and inference optimisation10% of the exam

   • Design inference services that meet latency, throughput and cost targets under variable load.
   • Scope: Online, batch and streaming inference, serving stacks (KServe, Triton, Ray Serve, BentoML), dynamic batching, quantisation and distillation, ONNX/TensorRT, autoscaling and latency/throughput/cost trade-offs.
6. 06Distributed training and GPU infrastructure operations6% of the exam

   • Evaluate and operate distributed training infrastructure for efficiency, reliability and cost.
   • Scope: Data, tensor and pipeline parallelism, FSDP and DeepSpeed, Kubernetes GPU scheduling, Kueue/Volcano, spot capacity and checkpointing, accelerator utilisation diagnostics and capacity planning.
7. 07LLMOps: serving, evaluation and RAG pipeline operations8% of the exam

   • Design and operate LLM and RAG systems with systematic evaluation, versioning and cost control.
   • Scope: LLM serving (vLLM, TGI), KV-cache and paged attention, prompt and adapter versioning, RAG index pipelines and vector stores, offline and online LLM evaluation, guardrails, token cost and latency management.
8. 08Production monitoring, drift detection and model observability10% of the exam

   • Diagnose model degradation in production and design monitoring that triggers timely, justified intervention.
   • Scope: Data and concept drift detection (PSI, KS, embedding drift), delayed-label performance estimation, ML-specific SLOs, alerting design, observability tooling (Evidently, WhyLabs, Arize) and retraining triggers.
9. 09Continuous training and ML testing strategy10% of the exam

   • Design continuous training and testing systems that detect regressions before they reach production.
   • Scope: CI/CD/CT for ML, data validation tests, model behavioural and invariance testing, fairness and robustness checks, ML Test Score rubric, automated retraining with rollback and feedback-loop risks.
10. 10ML supply-chain and model security6% of the exam

    • Evaluate ML systems for supply-chain and adversarial risks and implement layered security controls.
    • Scope: OWASP Top 10 for LLM Applications and ML Security Top 10, MITRE ATLAS, model serialisation risks (pickle vs safetensors), dependency and model provenance, secrets, poisoning, extraction and prompt-injection controls.
11. 11AI governance, regulatory compliance and auditability9% of the exam

    • Lead MLOps practices that produce the evidence and controls required for AI regulatory compliance.
    • Scope: EU AI Act risk tiers and obligations, ISO/IEC 42001 AI management systems, NIST AI RMF, technical documentation and logging requirements, human oversight, model risk management and audit evidence.
12. 12FinOps and reliability engineering for ML workloads9% of the exam

    • Evaluate ML workloads for cost efficiency and reliability and lead improvements against defined targets.
    • Scope: Cost attribution for training and inference, GPU right-sizing and sharing (MIG, time-slicing), caching and model routing, incident response for ML systems, chaos testing and disaster recovery for model services.

Question types

Single choice 100%

Delivery

Online (remote proctored), Test centre, On-site (corporate)

Retakes

Up to three attempts within any 12-month period. A 14-day waiting period applies after a first unsuccessful attempt and 30 days after a second. Each retake requires a new booking and fee. Retakes draw a different randomised form from the question bank.

Provisional cut score — psychometric validation is required before live delivery. Blueprint v1.0.

1. 1

   Register

   Create a free account and start the programme.

2. 2

   Practice

   Online practice test (coming soon).

3. 3

   Training

   Optional

   Online course or classroom — or skip it.

4. 4

   Certification

   Independent exam and certification decision.

Training is optional and does not by itself confer certification. Find a training partner