Get in Touch
 Duration 14 hours

Course Outline

Deployment Strategies

  • Comparing hosted API versus self-hosted advantages and trade-offs.
  • Patterns for on-premises, edge, and air-gapped environments.
  • Single model versus managing a model portfolio.
  • Comparative analysis of cost per decision.

Hardware Sizing and Capacity Planning

  • Assessing memory footprint relative to parameter count.
  • CPU-only execution versus GPU acceleration.
  • Planning for concurrency and throughput.
  • Allowing headroom for increasing task lengths.

Serving Stacks

  • Evaluating vLLM and SGLang options.
  • Utilising llama.cpp and GGUF execution.
  • Using Ollama for local development workflows.
  • Selecting tools based on the specific shape of the decision workload.

Quantisation While Preserving Calibration

  • Precision formats and associated memory savings.
  • Measuring quality degradation and calibration loss.
  • Per-tensor and importance-aware quantisation variants.
  • Conducting acceptance testing post-quantisation.

Latency Engineering

  • Analysing median versus tail latency.
  • Understanding latency growth as context length increases.
  • Implementing warm starts and model caching.
  • Batching multiple decision queries.

Edge and Constrained Environments

  • Running small decision models on CPU resources.
  • Deployment on Apple Silicon and embedded systems.
  • Offline operation strategies and update protocols.
  • Distributing weights in air-gapped scenarios.

Containerisation and Scaling

  • Designing images optimized for model serving.
  • Managing volumes and caches effectively.
  • Implementing horizontal scaling and routing.
  • Setting resource limits and sharing GPUs.

Observability

  • Tracking request, latency, and error metrics.
  • Monitoring confidence distributions.
  • Tracking decision outcomes and escalations.
  • Setting alerts for calibration drift.

Versioning and Rollout

  • Versioning models and adapters.
  • Deploying via shadow and canary methods.
  • Using pinned model files for audit purposes.
  • Establishing rollback procedures.

Security and Multi-Tenancy

  • Implementing authentication and rate limiting.
  • Ensuring tenant isolation for prompts and logs.
  • Managing network isolation and egress controls.
  • Maintaining audit logs for decisions.

Requirements

  • Proficiency in Linux administration and container technologies.
  • Practical experience with model inference servers and GPU resource management.
  • A solid grasp of REST APIs, load balancing strategies, and system monitoring.

Target Audience

  • MLOps and platform engineers managing inference infrastructure.
  • SRE teams responsible for supporting AI services.
  • AI engineers implementing on-premise decision model deployments.
  • Architects designing secure, air-gapped, or regulatory-compliant AI solutions.

Upcoming Courses

Related Categories