Course Objective & Scope
Many enterprise AI tasks—document classification, field extraction, PII detection, intent routing and templated summarisation—can be solved far faster and cheaper with well-chosen 1B–14B Small Language Models (SLMs) than with massive general-purpose models. This course examines the strengths and limits of SLMs across the Llama, Qwen, Gemma, Phi and Ministral families.
Participants generate verified training data from a larger teacher model to specialise an SLM for a narrow business problem, then benchmark the SLM against a frontier API on accuracy, latency and unit cost.
Who Should Attend?
- AI engineers replacing high-volume frontier API calls with faster, cheaper models
- Data scientists working on task-specific model adaptation and synthetic data generation
- Technical teams building on-premises solutions or working within tight GPU budgets
- Prerequisite: Python and basic experience building LLM applications
Key Takeaways
- Assess whether a task fits an SLM based on reasoning depth and context requirements
- Select 1B–14B open models by Turkish capability, licence and hardware footprint
- Improve SLM output consistency using constrained decoding and targeted few-shot prompts
- Apply task distillation by curating verified datasets from a stronger teacher model
- Deliver a cost, latency and quality comparison report between an SLM and a frontier model
Curriculum
Why Small Language Models (SLMs)?
- Capability boundaries of the 1B–14B parameter band and ideal enterprise use cases
- Unit cost, data privacy and low-latency benefits in high-volume pipelines
The SLM Ecosystem and Model Selection
- Comparing small models across Qwen, Llama, Gemma, Phi and Ministral
- Multilingual performance, Turkish token efficiency and commercial licences
Prompting and Constrained Decoding for SLMs
- Decomposing complex instructions into narrow pipeline steps for smaller models
- Enforcing valid structured output with JSON Schema and grammar constraints
Knowledge Distillation and Task Adaptation
- Generating verified synthetic labels and rationales from a frontier teacher model
- Closing the quality gap on narrow tasks with LoRA/QLoRA fine-tuning
Quantization, Edge and Single-GPU Deployment
- Balancing quality and memory with 4-bit and 8-bit quantization (AWQ, GGUF, FP8)
- High-throughput serving on a single GPU or CPU environment
Hands-On: Migrating a Task from Frontier LLM to SLM
- Porting an internal classification or extraction workload to an SLM
- Reporting quality delta, speedup and total cost of ownership
Other Courses in This Category
-
Deploying and integrating open-weight LLMs (Llama, Qwen, Mistral, DeepSeek) on-premises or in private cloud as OpenAI-compatible APIs using vLLM, SGLang and Ollama, including GPU sizing and quantization.
- Live Course
- 6 modules
-
Engineering techniques to cut token spend and p95 latency in production LLM applications: prompt and semantic caching, Batch APIs, output length control, tiered models and unit economics.
- Live Course
- 6 modules
-
Hands-on programme on building modular LLM applications with modern LangChain (LCEL, ChatModel abstractions, prompt templates, structured outputs, tool calling, retrievers and LangSmith tracing).
- Live Course
- 6 modules
-
Building type-safe LLM, RAG, tool-calling and MCP integrations directly in enterprise Java (Spring AI, LangChain4j) and .NET (Microsoft.Extensions.AI, Semantic Kernel) stacks without switching to Python.
- Live Course
- 6 modules
-
Hands-on programme for building enterprise generative AI solutions on AWS using the Amazon Bedrock Converse API, Knowledge Bases, Agents, Guardrails, IAM/VPC security and cost controls.
- Live Course
- 6 modules
-
Building enterprise AI applications on Azure AI Foundry and Azure OpenAI Service with GPT and reasoning models, Azure AI Search, Entra ID authentication, Private Endpoints, Content Safety and PTU/quota management.
- Live Course
- 6 modules
Get in touch
