AI0-001 Practice Exam 59 — 20 Questions

Question 1

An MLOps engineer is building a feature pipeline for a recommendation model. Features must be served to the online model with single-digit millisecond latency, while the same feature definitions must also be usable by offline training jobs to prevent training-serving skew. Which component of a feature store architecture directly satisfies the low-latency serving requirement?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

The online store backed by a low-latency key-value database. A feature store splits storage into an offline store for training and an online store for serving. The online store uses a low-latency key-value database so the model can retrieve the latest feature values in single-digit milliseconds. Because both paths are populated from the same registered feature definitions, the team also avoids training-serving skew, which is the second requirement in the scenario.

Answer

The feature registry that versions feature definitions

Answer

A scheduled batch job that materializes features every hour

Answer

The offline store backed by a data lakehouse

Question 2

A research team is developing an AI system to predict patient outcomes from electronic health records. The team must ensure the system adheres to ethical AI principles. Which TWO practices best align with the principle of transparency and explainability? (Choose two.)

Accepted Answer

Document the model's intended use, limitations, and performance across demographic subgroups in a model card.. The two practices that best align with transparency and explainability are documenting model details in a model card and implementing feature attribution for individual predictions. A model card communicates intended use, limitations, and subgroup performance, while feature attribution explains specific outputs. Together, they enable stakeholders to understand and scrutinize the model, fostering trust and accountability.

Answer

Restrict access to the model's source code and training data to a small group of developers.

Answer

Focus solely on maximizing predictive accuracy, as ethical concerns are secondary to clinical outcomes.

Answer

Use a complex ensemble model and provide only the overall accuracy metric to clinicians.

Question 3

An AI platform team runs inference for an image classifier on a shared GPU node. Multiple model replicas currently load the full model weights into GPU memory independently, and the node runs out of GPU memory when a third replica starts. The team wants to serve more replicas per GPU without changing model accuracy. Which approach best addresses the constraint?

Accepted Answer

Serve the replicas from shared GPU memory using a model server that supports weight sharing or a runtime that loads the model once.. The exhaustion comes from duplicate copies of identical weights, so the fix is to load the weights once and let multiple replicas reference them. Model servers and optimized runtimes that support shared weights or a single loaded model instance cut per-replica memory to activations and cache, increasing density without touching precision or accuracy.

Answer

Pin each replica to a separate CUDA stream and rely on the driver's scheduler to time-slice GPU memory.

Answer

Convert the model to FP16 and run the replicas with a smaller batch size to fit more instances.

Answer

Enable CUDA Unified Memory so the driver pages weights between host RAM and GPU memory on demand.

Question 4

A team trains a decision tree on a customer churn dataset with 40 features. The unpruned tree reaches 100% accuracy on the training set but only 68% on a held-out validation set. The team wants to reduce this gap without changing the algorithm. Which action is most appropriate?

Accepted Answer

Apply post-pruning with cost-complexity pruning. A tree with perfect training accuracy and much lower validation accuracy is overfitting, so the fix is to constrain its complexity. Cost-complexity pruning is the standard, algorithm-preserving remedy: it prunes branches that contribute little to reducing impurity and selects the penalty strength by cross-validation. The other actions either worsen overfitting, address it only indirectly, or destroy the ability to detect it.

Answer

Increase the maximum tree depth

Answer

Remove the validation set and report training accuracy

Answer

Add more training examples without changing hyperparameters

Question 5

An AI platform team is building a feature store that feeds both offline training jobs and an online model that must return features within a few milliseconds. They are concerned that a feature computed one way during training could be computed differently at serving time. Which design choice best prevents this training-serving skew?

Accepted Answer

Define each feature once in a shared transformation definition that is executed by both the batch and online paths.. Training-serving skew arises when the same feature is computed by different logic in the offline and online environments. A shared transformation definition executed by both engines makes the computation identical by construction, and the online path simply materializes the result into a low-latency store, satisfying the millisecond requirement without duplicating logic.

Answer

Precompute all features nightly and have the online model read them from the same batch tables used for training.

Answer

Log raw request payloads during serving and retrain the model on that log so training data matches production inputs.

Answer

Compute features with separate code paths for batch training and online serving, and reconcile differences in periodic audits.

Question 6

A media company runs an on-premises inference cluster for an image-tagging model. The model is trained on-premises and deployed into a container image that is rebuilt nightly in the company's internal registry. Security policy forbids any outbound internet access from the cluster. Which deployment approach best fits these constraints?

Accepted Answer

Export the trained model to ONNX, then serve it with ONNX Runtime inside the container image pulled from the internal registry.. An air-gapped cluster needs a self-contained inference stack: the runtime, the model, and all dependencies must travel inside the container image from the internal registry. Exporting to ONNX produces a portable, framework-independent graph that ONNX Runtime can execute without contacting any external service, so nightly rebuilds remain deterministic and no outbound traffic is required.

Answer

Deploy the model to a managed cloud inference endpoint and route internal cluster traffic to it through the corporate VPN.

Answer

Store the model in a public Hugging Face repository and have the container download it at startup with a read-only token.

Answer

Use TensorFlow Serving with a remote model repository mounted over SMB from a partner network share.

Question 7

A hospital's AI triage assistant produces recommendations that clinicians frequently override. An operations review finds the model was trained on data from a different patient population than the one currently served. Which action most directly addresses the root cause of the low acceptance rate?

Accepted Answer

Retrain or fine-tune the model using data representative of the current patient population and revalidate its performance. The review identified a population mismatch between training data and the patients now being served, so the corrective action must close that gap. Retraining or fine-tuning on representative local data and revalidating performance targets the cause directly. Confidence displays, override documentation, and higher refresh rates are peripheral measures that do not change what the model has learned.

Answer

Increase the model's inference frequency so recommendations refresh more often during a shift

Answer

Mandate that clinicians document a reason each time they override a recommendation

Answer

Add a confidence score display so clinicians can see how certain the model is about each recommendation

Question 8

A data scientist is preparing a dataset for training a machine learning model. The dataset contains a mix of numerical and categorical features, and some features have high cardinality. The data scientist needs to apply appropriate encoding techniques to transform categorical variables into a format suitable for the model. Which TWO encoding methods are most appropriate for high-cardinality categorical features? (Choose two.)

Accepted Answer

Target encoding. Target encoding and frequency encoding are both effective for high-cardinality categorical features. Target encoding replaces categories with the mean target value, capturing predictive relationships, while frequency encoding replaces categories with their counts, reducing dimensionality without imposing order. One-hot encoding creates too many columns, label encoding imposes false order, and binary encoding may not capture target relationships as well. Therefore, target and frequency encoding are the most appropriate methods.

Answer

Label encoding

Answer

Binary encoding

Answer

One-hot encoding

Question 9

A media company is building a recommendation model from user clickstream logs. The raw data arrives as millions of small JSON files in an object store, and nightly training jobs currently take over ten hours because the training cluster reads thousands of tiny files per second. The team wants to reduce training time without changing the model or the underlying data values. Which data engineering approach is most appropriate?

Accepted Answer

Compact the small JSON files into larger columnar files such as Parquet and have the training pipeline read those instead.. The bottleneck is the small-file pattern in object storage, not compute capacity or parsing speed. Repacking the same records into fewer large columnar files cuts listing, open, and read overhead and enables column pruning and compression, so the training job moves the same data values with far less I/O. This addresses the root cause while preserving the model and data semantics.

Answer

Convert the JSON files to CSV because CSV parsing is faster than JSON parsing.

Answer

Increase the number of worker nodes in the training cluster so more files can be read in parallel.

Answer

Cache the JSON files on local SSD storage on each worker node before training begins.

Question 10

A small e-commerce company wants to implement a chatbot to handle customer inquiries about order status and returns. The company has limited historical chat data and wants a solution that can be deployed quickly without extensive training. Which type of AI solution is most appropriate?

Accepted Answer

Fine-tune a pre-trained language model on the company's limited chat data and product information.. Fine-tuning a pre-trained language model is the most appropriate because it uses transfer learning to adapt a general model to the company's domain with minimal data. It provides natural language understanding and can be deployed quickly. Other options either require excessive resources, lack flexibility, or are too complex for the scenario.

Answer

Train a custom large language model from scratch using the company's product catalog.

Answer

Use a rule-based chatbot with predefined scripts for common inquiries.

Answer

Implement a reinforcement learning agent that learns from customer interactions in real time.

Question 11

A media company stores thousands of hours of raw broadcast footage in a cloud object storage bucket. A data engineering team needs a training dataset that contains only the short clips where a goal is scored, so they must locate and extract those specific time ranges from the video files before training. Which technology should the team use to extract the required segments from the video objects?

Accepted Answer

Use FFmpeg to decode the video and cut the required time ranges into new clip files.. Trimming video to precise time ranges is a media-processing task, and FFmpeg is the purpose-built tool for demuxing, decoding, and re-encoding streams with start and end timestamps. Object storage and table formats organize files but cannot parse video containers, so the extraction must be performed by a codec-aware utility before the clips enter the training pipeline.

Answer

Use a data lakehouse table format with schema evolution to store the video files as Delta tables.

Answer

Use OpenCV's VideoCapture with a GPU-accelerated codec to retrain the object detection model directly on the raw footage.

Answer

Use Apache Parquet to store each frame as a column and filter on the goal timestamp column.

Question 12

A software vendor is developing an AI system that generates realistic images of people for use in marketing campaigns. The system will be sold to clients in the EU. Under the EU AI Act, which obligation applies to the vendor regarding the generated content?

Accepted Answer

The vendor must ensure that generated images are marked in a machine-readable format as artificially generated.. The EU AI Act requires that AI-generated content, including synthetic images, be marked in a machine-readable format to indicate its artificial origin. This is a transparency obligation for providers of such systems. The vendor must implement this marking. Other obligations like consent, database registration, and conformity assessment apply to different risk categories or legal frameworks.

Answer

The vendor must register the AI system in the EU database for high-risk systems.

Answer

The vendor must obtain explicit consent from all individuals whose likeness is used in the training data.

Answer

The vendor must conduct a conformity assessment before placing the system on the market.

Question 13

A logistics company runs a route-optimization model on a fleet of delivery vehicles. Each vehicle has an NVIDIA Jetson module with limited memory, and connectivity is unavailable for hours at a time. The team wants the smallest possible runtime footprint while still executing the trained graph on the GPU. Which approach best fits these constraints?

Accepted Answer

Convert the model to a TensorRT engine for the target GPU. TensorRT compiles a trained network into a GPU-specific engine, fusing operations and choosing tuned kernels for the Jetson hardware. The engine is compact, runs entirely on the device, and needs no network, so the route optimizer keeps working during hours offline. Full frameworks, cloud endpoints, and raw weight files each fail one of the footprint, offline, or GPU-execution requirements.

Answer

Deploy the full training framework on each vehicle

Answer

Store the model as a CSV of weights and load it at startup

Answer

Serve the model from a cloud endpoint over the cellular network

Question 14

A hospital's AI governance committee is reviewing a sepsis-prediction model before deployment. The model was trained on five years of historical ICU data in which patients who received early antibiotics had better outcomes, and the model learned to recommend antibiotics for nearly every patient with any fever. The committee wants to determine whether the model has learned a spurious correlation rather than a true clinical signal. Which action best evaluates this concern?

Accepted Answer

Perform a feature-ablation study, removing fever and antibiotic-administration variables, and measure the change in predictive performance and decision patterns.. Feature ablation is the most direct way to test whether a model depends on suspected spurious inputs. By removing fever and antibiotic-administration variables and observing changes in performance and recommendation behavior, the committee can determine whether the model's decisions hinge on those confounded features rather than on genuine clinical indicators of sepsis.

Answer

Increase the size of the test set by repartitioning the existing data and recompute the AUC.

Answer

Deploy the model in shadow mode and compare its alerts against clinician judgment for one month.

Answer

Retrain the model with a larger learning rate and compare training loss across epochs.

Question 15

A team is deploying a sentiment classifier and notices that the model outputs probabilities such as 0.83 for the positive class, but the actual positive rate among examples scored near 0.83 is only about 0.55. Stakeholders need the scores to reflect true likelihoods. Which action should the team take?

Accepted Answer

Apply a calibration method such as Platt scaling or isotonic regression to the model's output scores.. The scenario describes a reliability problem: scores near 0.83 correspond to an actual positive rate near 0.55, so the probabilities are overconfident. Calibration techniques such as Platt scaling and isotonic regression learn a mapping from raw scores to empirical frequencies using a held-out set, producing probabilities that better reflect true likelihoods without changing the model's ranking ability.

Answer

Replace the classifier with a k-nearest neighbors model, which produces inherently calibrated probabilities.

Answer

Raise the decision threshold from 0.5 to 0.83 so only highly confident examples are labeled positive.

Answer

Retrain the classifier with a lower learning rate and more epochs to improve probability estimates.

Question 16

A support team wants an AI assistant that can answer employee questions by retrieving passages from the company's internal policy documents and generating a response grounded in those passages. The documents change weekly. Which approach should the team implement?

Accepted Answer

Retrieval-augmented generation. Retrieval-augmented generation pairs a document retriever with a generative model, so the assistant answers from the latest policy passages instead of relying on memorized weights. This keeps responses current as documents change weekly, supports citation, and reduces hallucination. Fine-tuning, training from scratch, and ungrounded prompting all fail to keep pace with changing content or provide grounding.

Answer

Training a model from scratch on the policy corpus

Answer

Fine-tuning the base model on all policy documents

Answer

Prompting the model with only the user's question

Question 17

A hospital's radiology department wants to run a diagnostic imaging model inside its own data center because patient images cannot leave the premises. The team needs to manage model versions, roll back a bad deployment quickly, and keep an audit trail of which model version produced each prediction. Which approach best satisfies these requirements?

Accepted Answer

Serve the model from an internal model registry with versioned artifacts and log the model version with each prediction.. Keeping the model in an internal registry preserves data residency while giving the team immutable versions for promotion and rollback. Logging the model version with every prediction produces the traceability regulators and clinicians need, so a questionable result can be tied to a specific artifact rather than to an undifferentiated deployment.

Answer

Copy the model file onto each radiologist workstation and run inference locally with a desktop script.

Answer

Host the model on a public inference API and send only anonymized image hashes for scoring.

Answer

Store the model in a shared network folder and have the serving application load whichever file has the newest timestamp.

Question 18

A financial institution is designing an AI system to detect fraudulent transactions in real time. The system must process 10,000 transactions per second with sub-10 ms latency. The team plans to use a gradient boosting model. Which infrastructure component is most critical to meet the latency requirement?

Accepted Answer

A low-latency model serving framework that supports in-memory inference and batching.. Sub-10 ms latency requires a serving framework that minimizes overhead and efficiently handles concurrent requests. In-memory inference with optimized batching reduces per-request time, while other components like Kafka, databases, or Kubernetes add latency or do not directly affect inference speed. The serving layer is the bottleneck for meeting such strict latency.

Answer

A container orchestration platform like Kubernetes for auto-scaling.

Answer

A distributed streaming platform like Apache Kafka for data ingestion.

Answer

A GPU-accelerated database for storing transaction history.

Question 19

A retailer's recommendation service runs on a managed inference endpoint. During a flash sale, request volume triples and the endpoint's response time exceeds the acceptable threshold. The operations team must reduce latency quickly without retraining the model. Which action should they take first?

Accepted Answer

Increase the number of provisioned inference instances behind the endpoint so requests are distributed across more compute.. The described symptom is a capacity shortfall during a traffic spike, so the fastest effective remedy is horizontal scaling of the serving tier. Adding inference instances distributes load and cuts queueing delay without touching the model. Retraining, trimming logging, and extending timeouts either take too long or merely conceal the latency, leaving the underlying bottleneck in place.

Answer

Raise the client-side request timeout so slow responses are tolerated instead of failing.

Answer

Lower the endpoint's logging verbosity and disable request tracing to reduce per-request overhead.

Answer

Retrain the model with a smaller architecture so each inference completes faster.

Question 20

A hospital wants to run a natural language processing model that summarizes clinical notes. Because of patient privacy regulations, the data cannot leave the hospital's on-premises network, and there is no dedicated GPU available. Which deployment approach best fits these constraints?

Accepted Answer

Deploy a quantized open-source model on an on-premises CPU server using a runtime such as ONNX Runtime or llama.cpp.. Running a quantized open-source model locally with a CPU-optimized runtime keeps all clinical data inside the hospital network and avoids GPU requirements. Quantization shrinks model size and speeds inference enough for summarization on standard servers. Cloud APIs, remote hosting, and full-precision large-model training all violate either the privacy boundary or the hardware constraint.

Answer

Host the model in a colocation facility that is physically separate but connected by a dedicated VPN.

Answer

Fine-tune a large foundation model on the hospital's CPU servers and serve it with full FP32 precision.

Answer

Use a cloud-based LLM API and send de-identified notes for summarization.