Thank you for sending your enquiry! One of our team members will contact you shortly.
Thank you for sending your booking! One of our team members will contact you shortly.
Duration 7 hours
Course Outline
Transitioning from Text Generation to Typed Decisions
- Understanding the output structure of decision models.
- Contrasting system-one decisions with reasoning LLM processes.
- Analyzing the cost and latency characteristics of bounded decision-making.
The Landscape of Open-Weight Decision Models
- Overview of Strands Decider 2B from AWS Labs.
- Exploring the JEV and GEV open-weight model families.
- Introduction to Pplx-Decider, Liquid D1, and Clef.
- Comparing licensing terms and deployment variances.
Architectural Deep Dive
- Utilizing an LLM torso devoid of a standard language modelling head.
- Implementing a pointer head to score candidate options.
- Understanding the derivation of confidence scores.
Practical Implementation: Running Your First Decisions
- Setting up local environments on CPUs and consumer-grade GPUs.
- Crafting effective questions and structuring option sets.
- Processing multiple inquiries in a single inference pass.
- Managing thresholds and configuring fallback behaviors.
Evaluating Decision Quality
- Utilizing JevBench and other decision-specific benchmarks.
- Balancing accuracy against calibration performance.
- Interpreting Brier scores and the reliability of confidence metrics.
- Recognizing tasks where decision models are not the appropriate solution.
Requirements
- Fundamental understanding of large language models and prompt-based application development.
- Proficiency with command-line interfaces and Python scripting.
- No prior experience in machine learning training is necessary.
Target Audience
- AI engineers assessing decision models for integration into agent workflows.
- Software developers seeking to substitute unbounded LLM calls with structured, bounded decisions.
- Solution and platform architects designing robust agent architectures.
- Technical product managers evaluating decision model options for product strategy.
