Level 1 Questions (5-6 Marks)
Unit 1
Chapter 1
Name and briefly explain the three principles of agency.
Name and briefly explain the three principles of AI that helps an agent.
Explain the SPA cycle (the operating principle) of an agent.
What are the advantages and disadvantages of multi-agent system over the single-agent system?
What do you understand by agent orchestration? Name the three components that enable agent orchestration.
Name any six agentic AI frameworks used in the industry.
What benefits do agentic AI frameworks offer?
What ethical challenges do agentic AI systems face? How do you overcome them?
Mention key applications of agentic AI systems.
Name the challenges faced by agentic AI systems.
Distinguish between traditional AI, generative AI, and agentic AI.
Chapter 2
Name and explain the core layers of an Agentic AI System architecture.
Name and explain the support layers of an Agentic AI System architecture.
With a neat picture and an appropriate example, describe the reflection pattern of an agentic AI system.
With a neat picture and an appropriate example, describe the tool use pattern of an agentic AI system.
With a neat picture and an appropriate example, describe the planning pattern of an agentic AI system.
With a neat picture and an appropriate example, describe the ReAct pattern of an agentic AI system.
With a neat picture and an appropriate example, describe the ReWOO pattern of an agentic AI system.
Unit 2
Chapter 3
In your own words, describe the concept of retrieval augmented generation (RAG).
With an example, explain the concept of tool calling in LangChain.
Chapter 4
What are embeddings? How do they help RAG?
Distinguish between zero-shot and few-shot prompting techniques.
What are chains in LangChain? Explain with an example.
Unit 3
Chapter 5
What is RAG? What are its benefits?
With a picture an an appropriate example, describe a single-agent agentic RAG system.
With a picture an an appropriate example, describe a multi-agent agentic RAG system.
With a picture an an appropriate example, describe a hierarchical agentic RAG system.
With a picture an an appropriate example, describe a query-planning agentic RAG system.
With a picture an an appropriate example, describe a routing based agentic RAG system.
With a picture an an appropriate example, describe a tool-use agentic RAG system.
With a picture an an appropriate example, describe a ReAct agentic RAG system.
With a picture an an appropriate example, describe an adaptive agentic RAG system.
With a picture an an appropriate example, describe a dynamic planning and execution agentic RAG system.
Describe the process of building an agentic RAG system using LlamaIndex.
What are the benefits of integrating Cohere with LlamaIndex while building an agentic RAG system?
Name two techniques that Cohere contributes to enhance the performance of an agentic RAG system built using LlamaIndex.
Unit 4
Chapter 6
Name essential components of a multi-agent system.
Name four types of multi-agent architectures.
Compare the features of LangGraph and CrewAI technologies used for building multi-agent systems.
Unit 5
Chapter 7
Explain key features of AgentOps, LangSmith, and LangFuse.
What is OpenTelemetry? What are its benefits?
Briefly explain the two types of instrumentation supported by OpenTelemetry.
Level 2 Questions (8 Marks)
Unit 1
Chapter 1
With a neat picture, briefly describe the seven layers of an agentic AI ecosystem.
With a neat picture, describe the spectrum of AI agents.
Chapter 2
With a neat picture, describe the central processing flow between components of an agentic AI framework.
Unit 2
Chapter 3
With a neat diagram, identify components of the LangChain ecosystem and briefly describe each component.
With appropriate examples, describe LangChain integrations.
Chapter 4
Name popular text splitting strategies supported by LangChain.
With a neat picture, describe the vector similarity search process implementation with LangChain.
Describe the support available in LangChain to manage memory of agentic AI systems.
How does LangChain support tool use by AI agents? Describe with an example.
Unit 3
Chapter 5
With a neat picture explain the architecture of an Agentic RAG system.
Describe the classification of Agentic RAG systems.
What are the limitations of agentic RAG systems?
Unit 4
Chapter 6
With a neat picture, describe the Crew, Agent, Tasks, and Flow of CrewAI.
Unit 5
Chapter 7
None.
Level 3 Questions (8-10 Marks)
Unit 1
Chapter 1
None.
Chapter 2
With a neat picture, describe the architecture of an Agentic-AI system.
With a neat picture, describe the single-agent architecture of an agentic-AI system.
With a neat picture, describe the multi-agent architecture of an agentic-AI system.
Name any eight of the ten components of an agentic-AI framework.
Unit 2
Chapter 3
With an appropriate example, explain the operations of LangGraph application having multiple nodes.
Chapter 4
With a neat picture, describe the document processing pipeline of LangChain.
Unit 3
Chapter 5
Distinguish between traditional RAG and agentic RAG.
Describe the architecture of LlamaIndex for implementing agentic RAG.
Unit 4
Chapter 6
Name four types of multi-agent workflows.
Describe the four types of hand-off techniques used by multi-agent systems.
Briefly describe four ways in which responsibility, control and information are handed-over from one agent to another in a multi-agent system.
A multi speciality hospital wants to build an AI powered “Appointment & Triage Assistant” using CrewAI. The system must read a patient’s basic details (age, gender, existing conditions, past visits), interpret a short text description of current symptoms, check doctor availability and consultation modes (online or in person), suggest three suitable slots with doctors ranked by relevance, and draft a confirmation message for SMS or WhatsApp to the patient. Help the hospital by
(i) identifying the Crew (overall goal)
(ii) defining at least three Agents with their roles
(iii) listing the key Tasks for each agent
(iv) describing the Flow showing the execution order and handoffs between agents
(v) mentioning any shared memory or tools needed.
A technical institute wants an AI based “Hostel Room Allocation Planner” using CrewAI. The system must read student data (year, branch, distance from home, preferences like quiet floor or friends list), check room and hostel availability, generate fair allocation suggestions with a justification score (based on rules like seniority and special needs), and prepare an official allocation letter for each student. Help the institute by
(i) identifying the Crew (overall goal)
(ii) defining at least three Agents with their roles
(iii) listing the key Tasks for each agent
(iv) describing the Flow showing the execution order and handoffs between agents
(v) mentioning any shared memory or tools needed.
A college cultural committee plans to build an AI driven “Event Planning & Promotion Assistant” using CrewAI. The system must take an event idea (type, target audience, expected budget), check institute rules and available venues, generate a detailed event plan with tentative schedule and resource list, design a basic promotion plan with posters and social media captions, and finally draft an approval email to the faculty coordinator and principal. Help the committee by
(i) identifying the Crew (overall goal)
(ii) defining at least three Agents with their roles
(iii) listing the key Tasks for each agent
(iv) describing the Flow showing the execution order and handoffs between agents
(v) mentioning any shared memory or tools needed.
A department wants to create an AI “Assignment Evaluation Assistant” using CrewAI. The system must accept assignment submissions (PDF or text), read the rubric and marking scheme, perform an initial evaluation with marks and brief comments per question, flag doubtful cases for manual review, and generate a summary report for the course teacher with statistics on average, median, and common mistakes. Help the department by
(i) identifying the Crew (overall goal)
(ii) defining at least three Agents with their roles
(iii) listing the key Tasks for each agent
(iv) describing the Flow showing the execution order and handoffs between agents
(v) mentioning any shared memory or tools needed.
A group of MTech students wants to design an AI powered “Research Literature Review Companion” using CrewAI. The system must take a research topic description, fetch relevant papers from online sources or uploaded PDFs, extract key contributions and methods from each paper, cluster papers into themes, and generate a structured survey style summary with gaps and possible directions for future work. Help the group by
(i) identifying the Crew (overall goal)
(ii) defining at least three Agents with their roles
(iii) listing the key Tasks for each agent
(iv) describing the Flow showing the execution order and handoffs between agents
(v) mentioning any shared memory or tools needed.
The Training and Placement Cell wants an AI based “Placement Preparation Coach” using CrewAI. The system must read a student’s profile (branch, CGPA, targeted companies, skills, past internships), fetch typical job descriptions and previous placement data, design a customised preparation plan (topics, timelines, resources), generate a set of practice questions for aptitude and technical rounds, and draft a weekly progress email for the student. Help the Placement Cell by
(i) identifying the Crew (overall goal)
(ii) defining at least three Agents with their roles
(iii) listing the key Tasks for each agent
(iv) describing the Flow showing the execution order and handoffs between agents
(v) mentioning any shared memory or tools needed.
A fintech startup in India wants to build an AI “Personal Finance & Budget Advisor” using CrewAI. The system must read a user’s income, expenses, loan EMIs, savings goals and risk profile, categorise recent transactions from bank statements, suggest a monthly budget with recommended limits per category, propose simple investment options suitable for the user, and generate a short advisory report in simple English. Help the startup by
(i) identifying the Crew (overall goal)
(ii) defining at least three Agents with their roles
(iii) listing the key Tasks for each agent
(iv) describing the Flow showing the execution order and handoffs between agents
(v) mentioning any shared memory or tools needed.
An IT services company wants an AI driven “Recruitment Screening Assistant” using CrewAI. The system must read job descriptions, parse candidate resumes, match skills and experience against job requirements, rank candidates with scores and short justifications, and generate a summary email for the HR manager listing top candidates and suggested interview rounds. Help the company by
(i) identifying the Crew (overall goal)
(ii) defining at least three Agents with their roles
(iii) listing the key Tasks for each agent
(iv) describing the Flow showing the execution order and handoffs between agents
(v) mentioning any shared memory or tools needed.
A state agriculture department wants to develop an AI based “Crop Advisory Assistant” using CrewAI. The system must read farmer and field data (location, soil type, crop, farm size), fetch current weather and market price information, recommend suitable crop care actions (irrigation, fertiliser, pest control) for the coming week, and generate a simple message in the local language that can be sent to the farmer by SMS. Help the department by
(i) identifying the Crew (overall goal)
(ii) defining at least three Agents with their roles
(iii) listing the key Tasks for each agent
(iv) describing the Flow showing the execution order and handoffs between agents
(v) mentioning any shared memory or tools needed.
A university library wants to implement an AI powered “Library Recommendation & Overdue Manager” using CrewAI. The system must read student reading history, current courses, and interests, analyse the library catalogue and availability, recommend books or e resources with brief reasons, identify overdue books and generate polite reminder messages, and prepare a monthly usage summary for the librarian. Help the library by
(i) identifying the Crew (overall goal)
(ii) defining at least three Agents with their roles
(iii) listing the key Tasks for each agent
(iv) describing the Flow showing the execution order and handoffs between agents
(v) mentioning any shared memory or tools needed.
Unit 5
Chapter 7
What do you understand by AI Agent Observability? Why is it important?
Describe the four steps of monitoring an AI agent and generating analytics using AgentOps. Name the Python functions and decorators used in these steps.
A global e-commerce company is deploying a multilingual customer support LLM assistant across multiple regions. They require strict p95 latency SLAs, real time cost and token usage dashboards per region, alerting on hallucination spikes, advanced RBAC for different country teams, PII redaction for GDPR compliance, and support for tracing across several non LangChain microservices and third party APIs.
Which platform (AgentOps or LangSmith, or LangFuse) best fits this scenario? Justify briefly.
A health tech startup is building a clinical notes summarisation assistant used by doctors during consultations. They need HIPAA compliant logging with automatic PHI redaction, detailed tracing across backend services (not built with LangChain), fine grained RBAC for auditors and doctors, latency and error rate SLAs with alerting, and cost breakdown per hospital.
Which platform (AgentOps or LangSmith, or LangFuse) best fits this scenario? Justify briefly.
An enterprise is rolling out an internal LLM “Dev Copilot” integrated into their CI/CD pipelines and issue tracker. They require detailed request tracing through Git hooks, build servers and deployment scripts (no LangChain), real time monitoring of failure rates, cost dashboards by team, SSO based RBAC, and environment separation for dev, QA, and production.
Which platform (AgentOps or LangSmith, or LangFuse) best fits this scenario? Justify briefly.
An edtech company is launching an LLM based Q&A assistant in their learning app. They want to run A/B tests across different model prompts, monitor latency and correctness metrics, trace interactions through a custom backend (without LangChain) and mobile clients, get per experiment cost analytics, and configure alerts for quality regressions and budget overshoots.
Which platform (AgentOps or LangSmith, or LangFuse) best fits this scenario? Justify briefly.
A private bank is creating an FAQ and workflow assistant for customers and branch staff. They need strict access control with RBAC and audit logs, comprehensive PII redaction in logs, dashboards for costs per business unit, cross service tracing between on premises microservices and external credit score APIs (no LangChain), and alerts on abnormal error or denial patterns.
Which platform (AgentOps or LangSmith, or LangFuse) best fits this scenario? Justify briefly.
A logistics company is building an LLM assistant to help plan delivery routes using live traffic and warehouse data. Their system uses event driven microservices and streaming APIs instead of LangChain. They need end to end tracing across these services, monitoring of token usage and cost per client, dashboards for latency and timeouts, and alerting on integration failures with external mapping APIs.
Which platform (AgentOps or LangSmith, or LangFuse) best fits this scenario? Justify briefly.
A law firm network wants an LLM based legal document review assistant. They require detailed trace logs of every model call for audits, automatic redaction of client identifiers and sensitive clauses in logs, RBAC by case and client, stage wise environments (sandbox, review, production), and rich dashboards for usage and cost per practice area. Their codebase is custom and does not use LangChain.
Which platform (AgentOps or LangSmith, or LangFuse) best fits this scenario? Justify briefly.
A SaaS provider offers an LLM powered personalisation engine to multiple retail clients. They need multi tenant dashboards for latency, error rates, and token costs per client, tenant level RBAC, PII redaction in stored traces, alerts when a given tenant exceeds configured budget or error thresholds, and integration with their existing microservice based architecture without LangChain.
Which platform (AgentOps or LangSmith, or LangFuse) best fits this scenario? Justify briefly.
A consumer tech company is rolling out a voice based LLM assistant that interacts with many tools like calendars, email, and IoT devices through separate services. They want cross tool and cross service tracing, live dashboards for latency and tool failure rates, alerting on spikes in failed tool calls, redaction of email content in logs, and support for dev / canary / prod environments. The orchestration is custom coded, not LangChain.
Which platform (AgentOps or LangSmith, or LangFuse) best fits this scenario? Justify briefly.
A startup is building an LLM powered coding interview practice platform. It uses custom evaluation pipelines for code execution, test case generation, and feedback, implemented as separate microservices. They require tracing across these services, cost and latency dashboards per question set, alerts on sandbox failures or abnormal error spikes, and RBAC for content creators vs platform admins, with no dependency on LangChain.
Which platform (AgentOps or LangSmith, or LangFuse) best fits this scenario? Justify briefly.