AIML Rounds
Track 02

AI Engineering

How LLM applications are served, built and run, from inference to retrieval and agents.

You should know:Logistic RegressionBackpropagation Through a Network

LLM inference and serving

Module 1

Lessons

How LLMs run fast and cheap: memory, batching, caching, decoding and scale.

  1. Prefill vs DecodeConcept30 min
  2. LLM QuantizationConcept35 min
  3. KV Cache Management and PagedAttentionConcept20 min
  4. Continuous Batching and SchedulingConcept20 min
  5. Prefix and Prompt CachingConcept20 min
  6. Speculative DecodingConcept20 min
  7. Serving Metrics and SLOsConcept20 min
  8. Scaling Inference: Parallelism, Replicas and RoutingConcept25 min
  9. Sizing and Cost: Cost per Million TokensConcept20 min

Module 2

Scenarios

Practice questions and real-world incidents to diagnose. New ones are added over time.

  1. LLM Inference PracticeInterview35 min

Prompt engineering

Module 3

Lessons

Structuring prompts, getting reliable output and choosing the right tool.

  1. Prompt Engineering PatternsConcept20 min
  2. Structured Output and Constrained DecodingConcept20 min
  3. Prompt vs RAG vs Fine-Tuning: How to ChooseConcept20 min

RAG

Module 4

Lessons

Building retrieval-augmented generation, from basic pipelines to advanced retrieval, and how to evaluate it.

  1. RAG End to EndConcept25 min
  2. Advanced RAG: Hybrid Search, Re-ranking, Query Rewriting and GraphRAGConcept25 min
  3. Evaluating LLM AppsConcept25 min

Module 5

Scenarios

Practice questions and real-world incidents to diagnose. New ones are added over time.

  1. RAG: Wrong Answer, Where to Look FirstInterview8 min
  2. RAG at Scale: Slow and WrongInterview8 min

Agentic AI

Module 6

Lessons

The agent loop, tools, planning, memory, MCP, multi-agent systems, infrastructure, evaluation and safety.

  1. What Is an Agent? The ReAct LoopConcept20 min
  2. Tool and Function CallingConcept20 min
  3. Planning and ReflectionConcept20 min
  4. Memory and Context EngineeringConcept20 min
  5. MCP: The Model Context ProtocolConcept20 min
  6. Multi-Agent SystemsConcept20 min
  7. Agent Infrastructure: Durable Execution, Retries and BudgetsConcept20 min
  8. Evaluating AgentsConcept20 min
  9. Agent Safety: Permissions, Sandboxing and the Lethal TrifectaConcept25 min

Module 7

Scenarios

Practice questions and real-world incidents to diagnose. New ones are added over time.

  1. Agentic AI PracticeInterview35 min
  2. Agent Scenarios: Set 1Interview35 min

MLOps

Module 8

Lessons

Skew, serving modes, deployment, reproducibility, drift and retraining.

  1. Training/Serving Skew and Feature StoresConcept20 min
  2. Batch vs Online vs Streaming InferenceConcept20 min
  3. Deployment Patterns: Shadow, Canary, A/B and RollbackConcept20 min
  4. Experiment Tracking, Reproducibility and the Model RegistryConcept20 min
  5. Drift Detection and MonitoringConcept25 min
  6. Retraining Pipelines and Continuous TrainingConcept20 min

Module 9

Scenarios

Practice questions and real-world incidents to diagnose. New ones are added over time.

  1. MLOps PracticeInterview35 min

LLMOps

Module 10

Lessons

Versioned configurations and release gates, guardrails, cost and latency.

  1. LLMOps: What's DifferentConcept20 min
  2. Guardrails and Prompt InjectionConcept25 min
  3. Cost and Latency: Routing, Caching and BatchingConcept20 min

Module 11

Scenarios

Practice questions and real-world incidents to diagnose. New ones are added over time.

  1. GenAI App Scenarios: Set 1Interview35 min

Next track

Next track: System Design

How AI and ML systems are designed, from requirements to rollout.

Open System Design