Skip to main content Back Grokking Hub Sign up Create account

Designing Generative AI Systems

Architect production GenAI — LLM assistants, RAG, diffusion, and video generation — and defend your designs in senior AI interviews.

  • Intermediate
  • 82 lessons
  • 11 modules
  • English
  • 8h 44m

Designed and created by Anand Rochlani

$99
See all plans Buy this course

Course content

11 sections · 82 lectures · 1 free to preview

GenAI Foundations & the Interview Playbook 10 lectures
  1. 1. Welcome! What Even IS Generative AI? Preview Video
  2. 2. Generative Models — The Creative Machines Video
  3. 3. Why GenAI Took Over the World Video
  4. 4. The Hardware Arms Race — Scaling Laws & Compute Video
  5. 5. Your Secret Weapon — The ML System Design Framework Video
  6. 6. Framing Problems Like a Senior Engineer Video
  7. 7. Data — The Fuel That Powers Everything Video
  8. 8. Architecture Deep Dive — Transformers & Attention Video
  9. 9. Training at Scale — From Loss Functions to Distributed GPUs Video
  10. 10. Does It Actually Work? Evaluation Strategies Video
Design a Smart Email Autocomplete (Text Generation I) 10 lectures
  1. 1. Gmail Smart Compose — The Feature Billions Use Daily Video
  2. 2. Framing the Problem + Data Sources Video
  3. 3. Tokenization — Teaching Machines to Read Video
  4. 4. Transformer Variants — Encoder, Decoder, or Both? Video
  5. 5. Inside the Decoder — Embeddings & Positional Encoding Video
  6. 6. The Transformer Engine + Prediction Head Video
  7. 7. Training — Pretraining, Finetuning & Combining Inputs Video
  8. 8. Sampling — How the Model Writes Your Emails Video
  9. 9. Evaluation — Is It Actually Good? Video
  10. 10. The Full System — Triggers, Filters & Bias Guards Video
Design a Real-Time Translation Service (Text Generation II) 10 lectures
  1. 1. Google Translate — One Billion Users, 130 Languages Video
  2. 2. Why Encoder-Decoder Wins for Translation Video
  3. 3. Preparing Translation Data Video
  4. 4. BPE — The Tokenizer That Powers Translation Video
  5. 5. The Encoder — Understanding the Source Language Video
  6. 6. The Decoder + Cross-Attention — The Secret Sauce Video
  7. 7. Training Stage 1 — Pretraining with Masked Language Modeling Video
  8. 8. Training Stage 2 — Finetuning + Sampling Video
  9. 9. Evaluation — BLEU, ROUGE & METEOR Video
  10. 10. The Full System — Language Detector + Translation Service Video
Design a Conversational LLM Assistant (Text Generation III) 12 lectures
  1. 1. ChatGPT — 100 Million Users in 3 Months Video
  2. 2. Data Prep — Cleaning the Entire Internet Video
  3. 3. Positional Encoding — Absolute, Relative & RoPE Video
  4. 4. Training Stage 1 — Pretraining on Trillions of Tokens Video
  5. 5. Training Stage 2 — Supervised Finetuning (SFT) Video
  6. 6. RLHF Part 1 — Training the Reward Model Video
  7. 7. RLHF Part 2 — PPO & The Final Model Video
  8. 8. Sampling — Top-k, Top-p & Temperature Video
  9. 9. Evaluation — Benchmarks That Actually Matter Video
  10. 10. Safety Evaluation — Toxicity, Bias & Adversarial Attacks Video
  11. 11. The Full System — Inference Pipeline Video
  12. 12. Wrap-Up — The Complete ChatGPT Blueprint Video
Design an Image Captioning Service (Multimodal) 6 lectures
  1. 1. Image Captioning — Where Vision Meets Language Video
  2. 2. Preparing 400 Million Image-Caption Pairs Video
  3. 3. Image Encoders — CNN vs. Transformer Video
  4. 4. Positional Encoding for Images + Text Decoder Video
  5. 5. Training, Sampling & CIDEr Evaluation Video
  6. 6. The Full System & Wrap-Up Video
Design a RAG Document Assistant (Retrieval) 8 lectures
  1. 1. RAG — When LLMs Need External Knowledge Video
  2. 2. Finetuning vs. Prompt Engineering vs. RAG Video
  3. 3. Document Parsing — Turning PDFs into Data Video
  4. 4. Indexing — Vector-Based Retrieval at Scale Video
  5. 5. Nearest Neighbor Search — Exact vs. Approximate Video
  6. 6. Prompt Engineering — CoT, Few-Shot & Beyond Video
  7. 7. Evaluation — The RAG Triad Video
  8. 8. The Full System & Wrap-Up Video
Design a Face Synthesis Engine (Image Generation I — GANs) 7 lectures
  1. 1. Face Generation — The GAN Chapter Video
  2. 2. VAE vs. GAN vs. Autoregressive vs. Diffusion Video
  3. 3. Data Prep + GAN Architecture — The Generator Video
  4. 4. GAN Architecture — Discriminator + Adversarial Training Video
  5. 5. GAN Loss Functions & Training Challenges Video
  6. 6. Sampling + Evaluation — FID & Inception Score Video
  7. 7. The Full System & Wrap-Up Video
Design a High-Resolution Image Generator (Image Generation II — Diffusion) 4 lectures
  1. 1. High-Res Image Synthesis — The Autoregressive Approach Video
  2. 2. The Image Tokenizer — VQ-VAE Deep Dive Video
  3. 3. The Image Generator + Training Video
  4. 4. Sampling, Evaluation & System Design Video
Design a Text-to-Image Platform (Image Generation III) 5 lectures
  1. 1. Text-to-Image — The Diffusion Revolution Video
  2. 2. Data Prep + U-Net Architecture Video
  3. 3. DiT + The Diffusion Training Process Video
  4. 4. Sampling — CFG, DDIM & Generation Challenges Video
  5. 5. Evaluation — CLIPScore + The Full System Video
Design a Personalized Avatar Generator (Image Generation IV — Fine-Tuning) 4 lectures
  1. 1. Personalized Headshots — Teaching AI Your Face Video
  2. 2. Three Methods — Textual Inversion, DreamBooth & LoRA Video
  3. 3. DreamBooth Training + Sampling Video
  4. 4. Evaluation + The Full System Video
Design a Text-to-Video System (Video Generation) 6 lectures
  1. 1. Text-to-Video — From Sora to Movie Gen Video
  2. 2. Latent Diffusion Models — The Compression Trick Video
  3. 3. Data Prep — 100 Million Videos at Scale Video
  4. 4. Architecture — Extending DiT for Video Video
  5. 5. Training Challenges + Sampling Video
  6. 6. Evaluation + The Full System Video

Description

A capability-track masterclass in generative AI architecture. You start with the foundations — how generative models, Transformers, and training at scale actually work — then design eleven production systems end to end across four tracks: text generation (smart autocomplete, real-time translation, a conversational LLM assistant), retrieval (a RAG document assistant), multimodal understanding (image captioning), and generative media (GAN face synthesis, high-resolution diffusion, text-to-image, personalized avatars, and text-to-video). Every design is built the way a senior engineer defends it in an interview: requirements first, then data, model, evaluation, and the serving system around it.