NCA-AIIO - Essential AI Knowledge (38% of the exam) - Section 1.7
Describe software components related to the lifecycle of AI development and deployment.
Describe the software tools involved at each stage of the AI lifecycle, from data preparation and model training with frameworks like PyTorch, through experiment tracking and MLOps pipelines, to model serving at inference time. Recognise how NVIDIA tools such as TensorRT and Triton Inference Server fit into deployment and optimisation stages.
Practice question for this objective
lock_openFree sampleEssential AI Knowledgemedium
A data science team has finished cleaning and versioning their training dataset and wants to begin iterative model experimentation. Which lifecycle stage immediately follows data preparation, and what is the primary software concern at that stage?
AModel registry - storing approved model artefacts for controlled promotion to production
BInference serving - deploying a model endpoint so applications can request predictions in real time
CTraining framework - orchestrating compute, defining model architecture, and running gradient-based optimisation on the prepared datacheck_circle Correct
DMonitoring - tracking data drift and model performance degradation against production baselines
Identify the software stage that directly follows data preparation in the AI development lifecycle and its core function. The AI development lifecycle progresses through data preparation, training, experiment tracking and evaluation, model registry, inference serving, and then ongoing monitoring. A training framework such as PyTorch, TensorFlow, or JAX is the primary software component at the training stage: it defines the model graph, manages GPU/multi-node compute allocation, and executes forward and backward passes over the prepared dataset. Entering inference serving or monitoring before training completes skips critical lifecycle stages.
Why A is wrong: A model registry manages promotion of validated models, not the iterative experimentation that comes directly after data preparation. Registries operate after training and evaluation are complete.
Why B is wrong: Inference serving comes after a model is trained and registered. Running a serving layer on unprepared, untrained weights produces no useful predictions and is the wrong stage.
Why C is correct: After data is prepared and versioned the next stage is model training, where a framework such as PyTorch or TensorFlow defines the network, manages distributed compute, and executes the optimisation loop.
Why D is wrong: Monitoring is a post-deployment concern that tracks live model behaviour. There is no production model to monitor immediately after data preparation, so this stage does not follow data prep.
Answers that look right on this material and are not. Each one is a distractor from a different question in the NCA-AIIO bank for this domain.
Triton Inference Server, which applies graph rewrites and precision reduction to each loaded model at startup before accepting requests.
errorWhy it is wrong: Triton Inference Server manages model lifecycle and routes inference requests to multiple backends; it does not itself perform graph optimisation or precision calibration. Those tasks are handled by TensorRT before the optimised engine is loaded into Triton.
Data versioning platform, because tracking dataset snapshots also captures the hyperparameters that produced each run
errorWhy it is wrong: Data versioning tools such as DVC focus on dataset and pipeline lineage rather than per-run metric capture and hyperparameter comparison. They complement experiment tracking but do not replace it.
TensorRT, which converts each model into an engine and then exposes a multi-backend HTTP and gRPC endpoint with dynamic request batching.
errorWhy it is wrong: TensorRT is closely related because it produces optimised engines, but it is an optimisation and runtime library, not a multi-framework serving application with dynamic request batching and a shared endpoint.