Model Online / On-Site
Duration
Modules 5 modules

Course Objective & Scope

General-purpose models frequently break formatting rules in production, ignore negative constraints in system prompts or require thousands of prompt tokens on every request just to maintain corporate tone. Instruction tuning teaches a model to internalise complex instructions, domain constraints and strict output schemas.

Hands-on labs cover designing task taxonomies, diversifying system prompts to prevent overfitting, scaling instruction complexity with Evol-Instruct, training for strict JSON/schema adherence, adding refusal and clarification examples, and evaluating adherence with IFEval-style programmatic verifiers. Participants train and verify an instruction-tuned model tailored to enterprise rules.

Who Should Attend?

  • AI engineers building enterprise assistants and workflow automation models
  • Data scientists training models for structured extraction, classification and summarisation
  • Application developers aiming to shrink bloated system prompts while improving output consistency
  • Prerequisite: Python and basic familiarity with LLMs and SFT

Key Takeaways

  • Design instruction datasets combining format, length, tone and negative constraints
  • Vary system prompts during training so the model generalises to new instructions instead of memorising one prompt
  • Achieve high-reliability JSON, XML and schema compliance without heavy runtime retries
  • Train calibrated refusal behaviour for out-of-scope requests and clarification questions for ambiguous inputs
  • Measure instruction-following precision using IFEval-style programmatic rule verifiers

Curriculum

Goals of Instruction Tuning and Behavioural Alignment
  • Teaching instruction-following skills versus memorising domain facts
  • Internalising fixed corporate rules (prompt compression) versus following dynamic runtime instructions
Instruction Taxonomies and Evol-Instruct Complexity Scaling
  • Balancing single-task specialisation with multi-task instruction diversity
  • In-depth (adding constraints, deepening reasoning) and in-breadth instruction evolution
  • Representing positive formatting rules alongside negative constraints
Structured Outputs, Multi-Turn Clarification and Refusals
  • Training for near-100% compliance on nested JSON schemas and structured tables
  • Teaching models to ask clarifying questions on ambiguous inputs rather than guessing
  • Calibrating policy refusals without triggering false-positive over-refusals
Verifiable Instruction Evaluation (The IFEval Approach)
  • Building code-verifiable constraint test suites independent of LLM judges
  • Measuring prompt-level and instruction-level strict and loose accuracy
Workshop: Instruction Tuning an Enterprise Assistant
  • Curating and training on an instruction dataset encoding corporate style and output rules
  • Benchmarking constraint adherence before and after instruction tuning

Other Courses in This Category

  • Preference Optimization: DPO, ORPO, KTO and LLM Alignment

    Advanced course on post-SFT alignment using DPO, IPO, KTO, ORPO and SimPO: building chosen/rejected preference datasets, tuning beta and KL constraints, and avoiding length bias and reward hacking.

    LLM Customisation & Model Adaptation Corporate
    Live Course
    5 modules
    View
  • Synthetic Data Generation: Data Synthesis and Distillation for LLM Training

    Hands-on course on generating high-diversity SFT and preference datasets in low-resource enterprise domains using teacher LLMs, Evol-Instruct, persona matrices, Distilabel and verification filters.

    LLM Customisation & Model Adaptation Corporate
    Live Course
    5 modules
    View
  • Training Data Design: Dataset Architecture for LLM Fine-Tuning

    Hands-on course on designing high-leverage SFT and multi-turn conversation datasets: the LIMA principle, task mixture ratios, expert annotation guidelines, schema design and dataset versioning.

    LLM Customisation & Model Adaptation Corporate
    Live Course
    5 modules
    View
  • Model Evaluation and Fine-Tuning Evals for Custom LLMs

    Course on evaluating fine-tuned LLMs for task accuracy, catastrophic forgetting regressions, hallucination and safety drift using lm-evaluation-harness, custom task suites and pairwise arena comparisons.

    LLM Customisation & Model Adaptation Corporate
    Live Course
    5 modules
    View
  • Open-Weight Model Adaptation: Customising Llama, Qwen, Mistral and Gemma

    Course on adapting open-weight models—Llama, Qwen, Mistral, Gemma and DeepSeek—for enterprise use: architectural differences, custom tokens, post-training quantisation (AWQ/GGUF/FP8) and self-hosted serving.

    LLM Customisation & Model Adaptation Corporate
    Live Course
    5 modules
    View
  • Turkish Language Models: Tokenization, Evaluation and Adaptation

    Course on adapting LLMs for Turkish: morphological challenges in BPE tokenization, vocabulary extension, continual pre-training, Turkish evaluation benchmarks and cultural/regulatory alignment.

    LLM Customisation & Model Adaptation Corporate
    Live Course
    5 modules
    View

Get in touch