Skip to main content

Retrieval சரியா work ஆகுதான்னு check பண்ணுங்க

ஒரு சின்ன test set வச்சு, recall at k, context precision மாதிரி metrics-ல retrieval-ஐ அளக்க முடியும்.

Intermediate 2 · RAG builderabout 3 hours7 sections

சுருக்கமா சொன்னா

கேள்விகளும் அதுக்கு பதில் இருக்குற passages-உம் அடங்குன ஒரு சின்ன test set வச்சு retrieval-ஐ check பண்ணுங்க. அப்புறம் recall at k அளக்கணும்: சரியான passage top k results-ல எவ்வளவு தடவை வருதுன்னு பாக்குறது.

இது ஏன் முக்கியம்

RAG answer தப்பா வந்தா, பெரும்பாலும் காரணம் model இல்லை, retrieval தான். Test set எது காரணம்னு சொல்லும். Improvement வந்துச்சுன்னு நம்பிக்கையில சொல்லாம, நிரூபிக்கவும் உதவும்.

Sign up free to read the whole lesson

The rest of this lesson: 7 parts, video lessons, the best course, practice and a quick check. It is free, takes a few seconds and needs no card.

Accounts are for learners aged 18 and over. Under 18? The quick answer and the first part of every lesson stay open: browse lessons by topic.

Lessons on AI Engineer0 of 70 done
  1. 1What AI, machine learning and a language model are
  2. 2AI careers in India
  3. 3Choosing your path
  4. 4Set up Python and run your first file
  5. 5Variables, loops and functions in Python
  6. 6Write your own functions
  7. 7Lists and dictionaries
  8. 8Files and errors
  9. 9The terminal and Git
  10. 10Packages and virtual environments
  11. 11Call a web API and read JSON
  12. 12Tables with pandas
  13. 13Maths for machine learning
  14. 14Statistics for machine learning
  15. 15Regression
  16. 16Classification
  17. 17Trees and gradient boosting
  18. 18Cross-validation and leakage
  19. 19Choosing a metric
  20. 20Neural networks
  21. 21PyTorch
  22. 22Transformers
  23. 23How a chat model is made: pre-training, SFT and preference tuning
  24. 24AI in Indian languages: tokens, quality gaps and what to use
  25. 25Model size and scaling laws: parameters, training data and compute
  26. 26Perplexity and cross-entropy: how language models are measured
  27. 27Call an LLM API from code
  28. 28System prompts and prompt design
  29. 29Tokens and the context window
  30. 30How a model picks its next word: temperature, top-p and top-k
  31. 31Get structured JSON output
  32. 32Errors, retries and cost
  33. 33Embeddings
  34. 34Chunk documents for retrieval
  35. 35Vector search
  36. 36Build a RAG app
  37. 37Check whether retrieval works
  38. 38Tool calling
  39. 39The agent loop
  40. 40Test sets and evals
  41. 41LLM judges and their limits
  42. 42Choosing a model: benchmarks, leaderboards and your own tests
  43. 43Scoring open-ended answers: exact checks, similarity and side-by-side comparison
  44. 44Catching hallucinations: why models make things up and how to detect it
  45. 45Regression tests for prompts
  46. 46Defend against prompt injection
  47. 47Docker and Compose
  48. 48REST APIs with FastAPI
  49. 49Serving a model
  50. 50Deploy and monitor an LLM app
  51. 51Monitoring and drift
  52. 52Caching and speed
  53. 53Inside LLM inference: prefill, decode and the KV cache
  54. 54Grow an AI app in layers: context, guardrails, routing, caching and agents
  55. 55Running models on many GPUs: replicas, tensor and pipeline parallelism
  56. 56Cost control
  57. 57Security and privacy
  58. 58Advanced harness engineering
  59. 59Harnesses for long-running agents
  60. 60How fine-tuning works: data, training and what can go wrong
  61. 61LoRA, QLoRA and the GPU memory maths
  62. 62Fine-tuning in practice: tools, settings and reading the loss curve
  63. 63Model merging: combining fine-tuned models into one
  64. 64Synthetic data and distillation: training with model-made data
  65. 65Building a training dataset: quality, coverage and quantity
  66. 66Reasoning models and test-time compute: when to let the model think longer
  67. 67Prompting, RAG or fine-tuning
  68. 68Design an AI system
  69. 69Quality, cost and speed trade-offs
  70. 70The design interview

On these roadmaps: AI Engineer, Generative AI & LLM Engineer, Forward Deployed Engineer (FDE)

Retrieval and RAGIntermediate 2: RAG builderVideo lessonsRecall at kGolden setRetrievalFaithfulness

Track

Today's AI turns are used

Lessons, roadmaps, videos, practice cards and the code playground stay free with your account, with no limit.