IBM AI Engineer interview questions

Continued pretraining adds domain knowledge that fine-tuning cannot, yet done carelessly it wrecks general ability. The signal is the replay ratio, learning-rate rewarming, and how you measure forgetting.

The staged recipe is the easy half. The half that sinks real projects is data licensing, an evaluation bar that needs actual domain experts, and knowing which parts of the problem fine-tuning cannot solve at all.

A from-scratch staple that checks whether you grasp the conditional-independence assumption and why you work in log space. The signal is fitting per-class means and variances, summing log-likelihoods, and adding the log prior. Here is the implementation.

The 'naive' independence assumption is wrong on real data, yet the classifier remains a strong baseline for text. The interesting answer explains why classification survives a broken assumption, plus smoothing and the variants. Here it is.

Fairness has no single number, and a well-known result proves the main three cannot hold together. What interviewers reward is defining each metric precisely and explaining the impossibility rather than picking one blindly. Here is the answer.

Healthcare is the highest-stakes AI domain: a wrong answer can injure a patient, and regulators are paying attention. The signal is designing decision support with grounding, human oversight, and guardrails, never an autonomous diagnoser. Here is the design.

Some questions need precise multi-hop facts that vector retrieval cannot chain. The signal is entity linking plus query translation over a graph (or GraphRAG), and knowing when the graph is worth its maintenance cost. Here is the design.

When a model regresses or an auditor asks 'how was this built,' guessing is a failing answer. What matters is tracking the entire chain so any model can be reproduced and any past decision explained. Here is the chain.

Promotion is the easy part. Governance covers who may ship what, on whose sign-off, with what documented, and how a model gets retired. Here is the control plane an auditor actually asks for.

Enterprise AI deals turn on governance, and engineers who can speak it stand out. What lands is treating it as documentation, accountability, and risk-tiered controls across the lifecycle rather than a legal afterthought.

Regulators and enterprises increasingly require explanations for AI decisions, and the two terms get thrown around loosely. What lands is separating inherently interpretable models from post-hoc explanations and picking the right technique for the stakes.

The NIST AI RMF is the most-cited voluntary framework for managing AI risk, and enterprises ask about it. What lands is treating it as a continuous process built on four functions rather than a compliance checklist. Here is the answer.

Model cards and datasheets are the standard responsible-AI documentation artifacts, and regulators and buyers increasingly demand them. The strong answer knows what each one covers and grasps that the whole purpose is honest disclosure of limitations, not marketing. Here is the answer.

The EU AI Act sorts obligations by risk, making that classification a first-order design decision rather than a legal footnote. The signal is knowing the four tiers and what high-risk systems must prove before they ship.

Audit logging is the backbone of AI accountability, yet logging everything blindly turns your logs into the biggest privacy liability you own. The signal is what to log so any decision can be replayed, and how to do it without hoarding PII.

Every third-party model, dataset, and library widens the attack surface: backdoors, poisoning, arbitrary code on load, license landmines. The signal is treating models and data as supply-chain artifacts that need provenance and vetting. Here is the answer.

For consequential decisions, 'the model said so' will not satisfy a regulator. Adverse-action notices and the right to explanation limit which model you can even ship. Here is the governance view.

ISO 42001 is the certifiable AI management-system standard, the AI counterpart to ISO 27001. The signal is knowing it is an auditable management system, not a technical control list, and how it sits alongside NIST and the EU AI Act. Here is the answer.

A fairness audit is a structured process: pick protected attributes, choose metrics that fit the harm, measure disaggregated performance, and document findings. The signal is knowing the metrics conflict and which one the use case requires. Here is the answer.

The EU AI Act hands different duties to high-risk deployers, GPAI model providers, and limited-risk systems. The signal is naming the specific obligations per tier, not merely reciting that risk tiers exist. Here is the answer.

Dropping the protected attribute is the answer most candidates give, and it accomplishes nothing: the attribute is reconstructible from the features you kept. Here is how to actually locate the proxies and what removing them costs you.

Single-axis fairness audits are just another form of averaging, and they average away the exact group that is being harmed. The hard part is not noticing that: it is handling the combinatorial blowup, the tiny cells, and the multiple-testing problem without chasing ghosts.

Distributed, cross-cultural work is the norm for customer-facing AI roles. Interviewers want to see deliberate async habits and cultural awareness, not heroics. Here is what scores.