Fine-tuning & RL for LLMs: Intro to Post-training

What you actually do

• Module 1: Post-Training Overview : Conversation between Sharon Zhou and Andrew Ng Video ・ 10m Background Video ・ 5m Where post-training (fine-tuning and RL) fits into LLM training Video ・ 6m Intuitions behind fine-tuning and RL Video ・ 4m Key components to making fine-tuning and RL work Video ・ 10m Post-training example: Reasoning Video ・ 5m Post-training example: Safety and security (RLAIF) Video ・ 4m Post-training in the wild Video ・ 4m Module 1: Quiz Graded ・Quiz ・ 30m Module 1: Graded Lab Graded ・Code Assignment ・ 1h Join the DeepLearning.AI Forum to ask questions, get support, or share amazing ideas! Reading ・ 5m Module 1 Lecture Notes Reading ・ 1m · Source
• Module 2: Core techniques in Fine-Tuning and RL : Data: What you need and how to prepare it Video ・ 7m Data: Tokens for models to read/write Data Video ・ 10m Fine-tuning math: Loss, gradients, weight updates (Part 1) Video ・ 6m Fine-tuning math: Loss, gradients, weight updates (Part 2) Video ・ 5m Fine-tuning: Hyperparameters & hyperparameter tuning (Part 1) Video ・ 7m Fine-tuning: Hyperparameters & hyperparameter tuning (Part 2) Video ・ 4m Module 2 Graded Lab: 1 Graded ・Code Assignment ・ 1h Fine-tuning: Parameter efficient fine-tuning (PEFT) Video ・ 9m RL: Rewards and preference learning Video ・ 9m RL: Training objective and RLHF Video ・ 8m RL: PPO and GRPO Algorithms Video ・ 8m Module 2: Quiz Graded ・Quiz ・ 30m Module 2 Graded Lab: 2 Graded ・Code Assignment ・ 1h Module 2 Lecture Notes Reading ・ 1m · Source
• Module 3: Evaluation as the North Star : Why evals are the north star Video ・ 2m Evals for post-training: Test sets and metrics Video ・ 8m RL test environments and monitoring RL updates Video ・ 5m Reward hacking Video ・ 4m Error analysis: Why it matters Video ・ 2m Error analysis: Diagnosing errors & interventions Video ・ 4m Error analysis: errors → causes → fixes Reading ・ 10m How to invest in good evals Video ・ 5m Red Teaming: Real world failures Video ・ 4m Module 3: Quiz Graded ・Quiz ・ 30m Module 3: Graded Lab Graded ・Code Assignment ・ 1h Module 3 Lecture Notes Reading ・ 1m · Source
• Module 4: Data Driven Post-Training : How much data you need for post-training Video ・ 7m Data for fine-tuning Video ・ 6m Data for RL (Part 1) Video ・ 7m Data for RL (Part 2) Video ・ 4m Putting it together Video ・ 2m Synthetic data pipelines Video ・ 7m Template engineering Video ・ 4m Constitutional AI, revisited Video ・ 5m Balancing data and rewards Video ・ 5m Module 4: Quiz Graded ・Quiz ・ 30m Module 4: Graded Lab Graded ・Code Assignment ・ 1h Module 4 Lecture Notes Reading ・ 1m · Source
• Module 5: Production Considerations : A production post-training pipeline Video ・ 7m Agents Video ・ 9m RL promotion rules (go/no-go) Video ・ 6m Data-feedback flywheel Video ・ 5m Monitoring and observability Video ・ 4m Infrastructure (Part 1) Video ・ 4m Infrastructure (Part 2) Video ・ 7m Production-ready checklist Video ・ 3m Module 5: Quiz Graded ・Quiz ・ 30m Module 5: Graded Lab Graded ・Code Assignment ・ 1h Acknowledgments Reading ・ 1m Module 5 Lecture Notes Reading ・ 1m · Source

What the credential is

Credential type

DeepLearning.AI Certificate of Completion · www.deeplearning.ai / fine-tuning-and-reinforcement-learning-for-llms-intro-to-post-training

Credential

DeepLearning.AI Certificate of Completion · info.deeplearning.ai / draft-understanding-certificates-and-accomplishments-in-deeplearning.ai-courses

Issuer

DeepLearning.AI · info.deeplearning.ai / draft-understanding-certificates-and-accomplishments-in-deeplearning.ai-courses

Sharing

The provider describes manually adding certificates to LinkedIn or using an available Add to LinkedIn link. · info.deeplearning.ai / draft-understanding-certificates-and-accomplishments-in-deeplearning.ai-courses

Exam requirement

The outline lists 11 graded assignments. A separate comprehensive final examination and passing threshold are not stated. · www.deeplearning.ai / fine-tuning-and-reinforcement-learning-for-llms-intro-to-post-training

Credential scope

As checked on 15 October 2026, Pro members completing the course receive a DeepLearning.AI Certificate of Completion. Free users do not receive certificates or accomplishments. Passing thresholds, separate final-exam requirements, expiry, renewal and academic credit are not established. The course header marks its 11 graded assignments as Pro access. · info.deeplearning.ai / draft-understanding-certificates-and-accomplishments-in-deeplearning.ai-courses

Cost details

Membership and refund terms: As checked on 15 October 2026, direct Pro membership upgrades take effect immediately. After a downgrade, access to content and eligibility to earn certificates continue until the next billing cycle, and progress on started courses is saved for 180 days. This does not turn short-course Accomplishments into official certificates. The seven-day trial can be cancelled to avoid charges. After the trial, new payments, renewals and upgrades are non-refundable; early cancellation or downgrade receives no partial refund, with access until the next renewal. Incorrect charges or charges after a trial cancellation may qualify for a full refund after support review. Eligible refunds typically take 5–10 business days, depending on the bank. These are direct-platform terms, not Coursera terms. · Source · Source · Source · Source

Source differences retained

As checked on 15 October 2026, the official page gives differing information on the points below. Each reading is kept as found; none is reconciled.

• Language: As checked on 15 October 2026, the matching Course metadata states inLanguage="en" and availableLanguage=["en","jp","kr"]. English is retained; the additional literal codes do not establish teaching audio, subtitles or another language-option scope. · Source
• Duration: The visible course duration is 13h19m; same-course embedded courseWorkload PT1H differs. No numeric completion estimate is inferred. · Source

Who should skip it

• Free content access excludes certificates and Accomplishments; recognition requires Pro membership. The course header marks its 11 graded assignments as Pro access. · Official source · 2026-10-15
• Direct-platform certificates are issued by DeepLearning.AI; a Coursera-hosted route has separate credential and billing terms. · Official source · 2026-10-15

Alternatives in the registry

Related by topic or level.

• Finetuning with Llama : Check official price; 2 hours (provider headline; section and exam times listed separately) (duration checked 25 September 2026).
• Introduction to LLaMA : Check official price; 2 hours (provider headline; module and exam timings listed separately) (duration checked 27 September 2026).
• Efficient Inference with SGLang: Text and Image Generation : Free (As shown in the United States on 15 October 2026, the provider help page states Short Courses are free; Accomplishments require a Pro subscription. This zero is scoped to free content access, not a credential-inclusive completion total. The captured membership page does not expose a numeric price in its visible text. External API, cloud, compute and tool costs are not quantified.); 1h19m listed course duration (duration checked 15 October 2026).

Change log

• 2026-10-15: Initial source review from the dated US capture. · Source