Explain ML and LLM fundamentals

Read the full interview experience this question came from →

Quick Overview

This question evaluates understanding of core machine learning and LLM competencies including evaluation metrics (F1 vs accuracy), differences between classification and regression, fair benchmarking, inference pipeline bottlenecks and optimization, productionization of AI services, and modern LLM concepts such as the Transformer architecture, context engineering, retrieval-augmented generation, grounding, and guardrails. It is commonly asked in Machine Learning interviews to assess both conceptual understanding and practical application, probing the candidate's ability to reason about evaluation trade-offs, system-level performance constraints, and deployment considerations rather than only algorithmic detail.

Explain ML and LLM fundamentals

Company: Salesforce

Role: Machine Learning Engineer

Category: Machine Learning

Difficulty: medium

Interview Round: Onsite

You are interviewing for an AI Engineer role. Explain the following concepts and how they affect real systems: 1. What is F1 score, and when is it more useful than accuracy? 2. How do classification and regression differ in objective, output, and evaluation? 3. How would you benchmark an ML or LLM-based system fairly? 4. What are common bottlenecks in model inference pipelines, and how can you optimize latency and throughput? 5. How would you productionize an AI agent service? 6. For modern LLM applications, explain: - the Transformer architecture at a high level - context engineering - retrieval-augmented generation - grounding - guardrails Answer with practical examples and trade-offs.

Overview: This question evaluates understanding of core machine learning and LLM competencies including evaluation metrics (F1 vs accuracy), differences between classification and regression, fair benchmarking, inference pipeline bottlenecks and optimization, productionization of AI services, and modern LLM concepts such as the Transformer architecture, context engineering, retrieval-augmented generation, grounding, and guardrails. It is commonly asked in Machine Learning interviews to assess both conceptual understanding and practical application, probing the candidate's ability to reason about evaluation trade-offs, system-level performance constraints, and deployment considerations rather than only algorithmic detail.

Read the full Salesforce Machine Learning Engineer interview experience this question came from

|Home/Machine Learning/Salesforce
Salesforce logo
Salesforce
Jan 12, 2026
mediumMachine Learning EngineerOnsiteMachine Learning
11
0

You are interviewing for an AI Engineer role. Explain the following concepts and how they affect real systems:

  1. What is F1 score, and when is it more useful than accuracy?
  2. How do classification and regression differ in objective, output, and evaluation?
  3. How would you benchmark an ML or LLM-based system fairly?
  4. What are common bottlenecks in model inference pipelines, and how can you optimize latency and throughput?
  5. How would you productionize an AI agent service?
  6. For modern LLM applications, explain:
    • the Transformer architecture at a high level
    • context engineering
    • retrieval-augmented generation
    • grounding
    • guardrails

Answer with practical examples and trade-offs.

Loading comments...

Browse More Questions

More Machine Learning•More Salesforce•More Machine Learning Engineer•Salesforce Machine Learning Engineer•Salesforce Machine Learning•Machine Learning Engineer Machine Learning