Series

Fine-Tuning

Adapting models after pretraining: LoRA, SFT, DPO, RLHF, data size, epochs and cost.

38 episodes
  1. 1

    How Does Fine Tuning Work Anyway?

    Unlock the secrets of AI fine-tuning. Discover how your small dataset can shape a giant model.

    34:18 · 24 Nov 2025
  2. 2

    System Prompts vs. Fine-Tuning: Are We Building Solutions for Problems That Don't Exist?

    Are we over-engineering AI solutions? We dive into system prompts vs. fine-tuning and ask: Do you even need to fine-tune?

    29:45 · 11 Dec 2025
  3. 3

    System Prompts vs Fine-Tuning: When to Actually Train Your AI

    Prompt or fine-tune? We break down when to train your AI, from Shakespearean emails to law firm docs. Avoid unnecessary fine-tuning!

    23:36 · 16 Dec 2025
  4. 4

    Why Your AI Is a Yes-Man

    Ever wonder why AI is so polite? Herman and Corn dive into the mechanics of RLHF and how "niceness" gets baked into modern language models.

    26:33 · 29 Dec 2025
  5. 5

    The LoRA Revolution: Training AI for Personal Perspective

    Discover how to train LoRAs for character consistency and unique locations while avoiding common pitfalls like over-fitting and dataset bias.

    27:13 · 9 Feb 2026
  6. 6

    Is RLHF Lobotomizing AI? Why Guardrails Kill IQ

    Are safety guardrails making AI less intelligent? Explore the "alignment tax" and why corporate filters might be lobotomizing our best tools.

    30:09 · 13 Mar 2026
  7. 7

    Why We Still Fine-Tune in 2026

    Despite million-token context windows, fine-tuning remains essential. Here’s why behavior, not just facts, matters.

    29:23 · 2 Apr 2026
  8. 8

    Skip Fine-Tuning: Shape LLMs With Alignment Alone

    Can you build a personalized LLM by skipping traditional fine-tuning and using only post-training alignment methods like DPO and GRPO? We break dow...

    23:45 · 12 Apr 2026
  9. 9

    When AI Optimizes the Wrong Thing

    Discover how AI systems learn to optimize for rewards—and why they sometimes get it dangerously wrong.

    26:04 · 19 Apr 2026
  10. 10

    How to Bake Personality Into an LLM in 15 Minutes

    Fine-tune a model's personality with ~300 examples and a consumer GPU. SFT + DPO explained.

    25:41 · 27 Apr 2026
  11. 11

    Overfitting Is Not a Binary Condition

    Overfitting isn't binary. Learn the real triggers, the bias-variance tradeoff, and modern techniques to prevent it.

    33:59 · 29 Apr 2026
  12. 12

    How Unsloth Makes LLM Fine-Tuning 2x Faster

    Unsloth cuts memory usage by 50-70% and speeds up training 2.2x for models like Llama 3 and Mistral.

    30:40 · 29 Apr 2026
  13. 13

    Fine-Tuning DeepSeek for One Podcast

    Can a purpose-specific fine-tune fix a model's stubborn writing tics? We explore the practical engineering behind it.

    32:26 · 5 Jun 2026
  14. 14

    LoRA Isn’t Just for Image Generation

    LoRA lets you fine-tune an LLM’s behavior with a 50MB file. Here’s how it works and why it matters.

    30:19 · 9 Jun 2026
  15. 15

    Can a Podcast Train Itself?

    Can a podcast become a self-developing project? We explore RAG vs. fine-tuning for lore memory.

    23:47 · 23 Aug 2026
  16. 16

    Fine-Tune vs Prompt: Taming Hallucinated Deadlines

    Big models add fake deadlines. Small models miss tasks. Which wrongness can you build a system around?

    34:24 · 1 Sept 2026
  17. 17

    Masked Safety: How Post-Training Changes AI Behavior

    Safety mechanisms aren't erased in post-trained models—they're masked. Here's how that changes everything for military AI.

    26:49 · 3 Sept 2026
  18. 18

    Fine-Tuning vs From-Scratch for Minor Languages

    What 774 experiments and a new Armenian model reveal about the tradeoff between fluency and knowledge in low-resource languages.

    23:44 · 10 Sept 2026
  19. 19

    Fine-Tuning a Model on 100 Hand-Edited Answers

    You don't need 10,000 examples to make a model sound like you. The real number is closer to 100 — if the edits are opinionated.

    21:21 · 20 Sept 2026
  20. 20

    Adapters: 102KB That Reshapes a 403GB Model

    A 102KB adapter file changes how a 403GB base model behaves — without ever merging into it. Here's how model adapters actually work.

    27:03 · 21 Sept 2026
  21. 21

    Fine-Tuning at 4-Bit vs 16-Bit: What It Really Costs

    QLoRA cuts fine-tuning VRAM 15x and cost up to 85% — but you pay in training time, quality, and safety alignment.

    20:49 · 21 Sept 2026
  22. 22

    Editing vs. Note-Taking for AI Fine-Tunes

    Hand-editing a model's output gives three training signals at once. Writing notes gives one — and a weaker one at that.

    20:57 · 21 Sept 2026
  23. 23

    Keeping a Fine-Tune Alive Across Model Releases

    Daniel wants to fine-tune DeepSeek Flash 4.1 on edited podcast scripts. The hard part isn't training — it's surviving the next release.

    18:09 · 24 Sept 2026
  24. 24

    One Forward Pass: Building a Bounded Classifier

    No generation, no parsing — just label scores in a single pass. How to build a small encoder classifier that actually works.

    21:52 · 28 Sept 2026
  25. 25

    Serving Your Own Fine-Tuned Model in the Cloud

    You fine-tuned an open-weight model. Now how does anyone actually talk to it? Dedicated GPUs vs serverless inference, and the math that decides it.

    24:20 · 10 Oct 2026
  26. 26

    The Fine-Tuning Career: Titles, Skills, and What It Pays

    The job above the labeling layer exists — it just doesn't have one name. What it takes, what it pays, and how to sell it.

    25:54 · 13 Oct 2026
  27. 27

    Teaching a Model Your Voice: SFT vs DPO

    Three ways to teach a model your writing voice — and the evidence says one of them quietly makes things worse.

    21:34 · 13 Oct 2026
  28. 28

    How a Reward Model Actually Trains an LLM

    One number, no gradients, four models in the room. What a reward model actually does inside the RLHF loop.

    28:58 · 14 Oct 2026
  29. 29

    Keeping the Rollout Engine Fed: Inside RLHF Pipelines

    Most of an RLHF run is spent waiting on generation, not gradients. Here's how TRL, OpenRLHF, and verl actually differ.

    24:48 · 14 Oct 2026
  30. 30

    Blending Datasets vs Merging Weights for Fine-Tuning

    Merging weights happens in parameter space. Blending datasets happens in data space — and the messy part is worth it.

    20:56 · 16 Oct 2026
  31. 31

    How to Build an AI Preference Dataset

    Pairwise comparison, forced choice, or just "this one or that one" — the engineering behind collecting human preferences on AI outputs.

    22:35 · 19 Oct 2026
  32. 32

    What Actually Happens During Fine-Tuning

    Sixteen bytes per parameter in, two out — a walk through the forward pass, backward pass, and optimizer that turn a dataset into new weights.

    21:05 · 21 Oct 2026
  33. 33

    Fine-Tuning a Trillion-Parameter Model at Home

    What does it actually take to full fine-tune a trillion-parameter model? Sixteen terabytes of optimizer state, hundreds of GPUs, and a very big bill.

    21:24 · 22 Oct 2026
  34. 34

    Nobody Trains an Uncensored Model From Scratch

    Every "uncensored" model is a renovation, not a new build — and the political lean lives in the agreement layer, not the refusals.

    19:56 · 22 Oct 2026
  35. 35

    How Many Epochs Does Fine-Tuning Actually Need?

    Epochs are a receipt, not a plan. How to pick a training ceiling, read the curves, and know when to stop.

    20:20 · 23 Oct 2026
  36. 36

    Why Fine-Tuning Runs Can't Always Resume

    Eighty cents and twenty-five minutes per training run — so why does the checkpoint question still keep a tinkerer up at night?

    20:38 · 23 Oct 2026
  37. 37

    How Many Examples Does Fine-Tuning Actually Need?

    Everyone says "thousands." Nobody says how many for your task. Here's how to measure the number yourself.

    21:23 · 23 Oct 2026
  38. 38

    Training a LoRA to Write Like You

    A Humanizer model stripped every trace of the author. The fix isn't a better Humanizer — it's a LoRA trained on your own writing.

    21:40 · 23 Oct 2026