Model Online / On-Site
Duration
Modules 5 modules

Course Objective & Scope

One of the costliest mistakes in enterprise AI is running fine-tuning jobs to solve dynamic knowledge problems that belong in RAG—or, conversely, stuffing massive system prompts to force strict formatting and low latency where a small fine-tuned model excels. This course provides an evidence-based engineering framework for choosing the right adaptation path.

Participants compare in-context learning with prompt caching, RAG, LoRA fine-tuning, RAFT (Retrieval-Augmented Fine-Tuning) and model distillation on the same benchmark task. Each team leaves with a concrete adaptation decision matrix and staged roadmap for its own use cases.

Who Should Attend?

  • AI engineers and solution architects setting technical direction for LLM initiatives
  • Data science leads prioritising RAG versus model-training investments
  • AI product managers responsible for technical feasibility and roadmaps
  • Prerequisite: familiarity with LLM APIs, basic RAG and prompt engineering

Key Takeaways

  • Separate knowledge-injection requirements from behaviour, tone and schema-adaptation needs
  • Compare prompting, prompt caching, RAG and fine-tuning across accuracy, latency and token cost
  • Identify metric thresholds for combining RAG with domain-adapted models (RAFT)
  • Calculate the volume break-even point for distilling a frontier model into a small fine-tuned model
  • Build a reusable adaptation decision matrix for internal AI projects

Curriculum

The Adaptation Spectrum: Diagnosing the Problem Type
  • Separating missing facts (knowledge) from missing form, style or domain reasoning (behaviour)
  • Impact of data volatility, citation requirements and document-level access control
Context Engineering and Long-Context Limits
  • System prompts, dynamic few-shot selection and prompt-caching economics
  • Time-to-first-token (TTFT) and attention degradation in large context windows
RAG and Fine-Tuning: Complementary, Not Competing
  • Where RAG wins: mutable data, ACL enforcement and verifiable citations
  • Where fine-tuning wins: dense domain jargon, strict schema compliance, low latency and compact prompts
  • Hybrid architectures: fine-tuned embeddings, custom rerankers and RAFT
Model Distillation and Volume Break-Even
  • Distilling a frontier model + long prompt into a small fine-tuned open model (Llama, Qwen, Gemma)
  • Calculating training + serving TCO break-even points against daily request volume
Workshop: Enterprise Adaptation Decision Matrix
  • Evaluating three real organisational use cases through the decision framework
  • Drafting a staged validation roadmap (Prompt → RAG → Fine-Tuning)

Other Courses in This Category

  • Model and Data Readiness: Base Model Selection and Data Suitability

    Course on selecting open-weight or commercial base models by size, licence, tokenizer and hardware footprint, and auditing enterprise data volume, quality and privacy readiness before adaptation.

    LLM Customisation & Model Adaptation Corporate
    Live Course
    5 modules
    View
  • Quality, Cost and Risk Analysis in LLM Adaptation

    Course on modelling total cost of ownership (data curation, training and 24/7 serving), quality-versus-regression trade-offs and operational/regulatory risks for LLM customisation projects.

    LLM Customisation & Model Adaptation Corporate
    Live Course
    5 modules
    View
  • Baselines and Success Metrics: Pre-Adaptation Benchmarking for LLMs

    Course on building golden test sets, establishing zero/few-shot baselines, defining deterministic and LLM-as-a-judge metrics and setting acceptance thresholds before starting model adaptation.

    LLM Customisation & Model Adaptation Corporate
    Live Course
    5 modules
    View
  • Fine-Tuning Fundamentals: Introduction to Customising LLMs

    Foundational hands-on course on how LLM fine-tuning works: dataset formatting, chat templates, running a first adapter training job with Hugging Face TRL and Unsloth, and avoiding overfitting.

    LLM Customisation & Model Adaptation Corporate
    Live Course
    5 modules
    View
  • Advanced Fine-Tuning: Continual Pre-Training, Long Context and Model Merging

    Advanced course covering continual pre-training (CPT), RoPE long-context extension, multi-turn tool-calling fine-tuning, catastrophic forgetting mitigation and model merging (SLERP, TIES, DARE) with MergeKit.

    LLM Customisation & Model Adaptation Corporate
    Live Course
    5 modules
    View
  • Supervised Fine-Tuning (SFT): Engineering and Optimisation

    Deep-dive engineering course on Supervised Fine-Tuning (SFT): completion-only loss masking, sequence packing, FlashAttention-2, NEFTune, learning rate schedules and production workflows with TRL SFTTrainer and Axolotl.

    LLM Customisation & Model Adaptation Corporate
    Live Course
    5 modules
    View

Get in touch