GenAI & Agentic AI — Exam Prep¶
A study companion for the upGrad × IIT Kharagpur Certificate Course on Generative and Agentic AI.
This site turns twelve dense lectures into a study path you can actually finish — deep learn-guides for understanding, compressed notes for revision, a mock exam for practice, and a one-page cheat sheet for the day before.
Exam at a glance
- Format: 24 multiple-choice questions
- Marks: 4 per question (96 total)
- Duration: 90 minutes
- Negative marking: None — attempt every question
- Passing bar: typically 50% (aim for 70% with buffer)
Start here¶
-
New to the material?
Start with the deep learn-guides. Plain English explanations, analogies, worked examples, and 5 practice MCQs per lecture.
-
Already studied — need a refresh?
Compressed formulas, flashcards, and tables. Skip the story, keep the substance.
-
Want to test yourself?
Interactive 24-question mock exam with instant scoring, matching the real paper's difficulty.
-
Day before the exam?
One-page skim doc. Every formula, table, and trap from the sample paper.
What's in each lecture¶
| # | Topic | Key concepts |
|---|---|---|
| 1 | ML foundations | Precision / Recall / F1, confusion matrix, ML pipeline |
| 2 | Neural networks & word embeddings | CNN vs MLP, Word2Vec, LSTM shapes, architecture families |
| 3 | Transformers | Q/K/V attention, positional encoding |
| 4 | LLM decoding & APIs | Greedy/top-k/top-p, OpenAI Chat API, HF Inference |
| 5 | Structured outputs & evaluation | BLEU / ROUGE / METEOR / BERTScore |
| 6 | AI safety | CIA triad, prompt injection, jailbreaks |
| 7 | Advanced prompting | Zero-shot / Few-shot / CoT / ReAct / ToT |
| 8 | Prompt security & APE | Levenshtein distance, log-prob scoring |
| 9 | RAG fundamentals | Retrieval → augment → generate |
| 10 | Dense retrieval | Bi-encoder vs cross-encoder, two-stage retrieve→rerank |
| 11 | Vector indexing | Product Quantization math, IVF, HNSW, RBAC gateway |
| 12 | Agentic RAG | Observe-Think-Act loop, tool routing |
How this site is organised¶
The same 12 lectures appear in three depths so you can pick the one that matches your time and energy today:
- Learn — 800–2,000 lines per lecture. Plain-English concept walkthroughs. This is the primary reading.
- Revise — 200–400 lines per lecture. Formulas, tables, flashcards. Great once you already understand it.
- Cheat sheet — one page total. For the last hour before the exam.
Credits¶
Course content, lecture slides, and the sample paper are © upGrad × IIT Kharagpur and are not redistributed on this site. The study guides here are original explanatory material written by learners preparing for the same exam, released under CC BY 4.0.
Read the About page for attribution details and contribution guidelines.