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
- Nature and output of decision model responses.
- Contrast between System One decision processing and reasoning LLMs.
- Analysis of cost efficiency and latency characteristics in bounded decision scenarios.
The Ecosystem of Open-Weight Decision Models
- Overview of Strands Decider 2B by AWS Labs.
- Examination of the JEV and GEV open-weight model families.
- Insights into Pplx-Decider, Liquid D1, and Clef.
- Comparison of licensing structures and deployment variations.
Architectural Deep Dive into Decision Models
- Utilizing an LLM torso without a standard language modelling head.
- Scoring candidate options through a pointer head mechanism.
- Understanding the derivation of confidence scores.
Executing Initial Decision Workflows
- Setting up local environments on CPU and consumer-grade GPU hardware.
- Strategy for formulating questions and defining option sets.
- Techniques for processing multiple questions in a single inference pass.
- Configuring thresholds and defining fallback behaviors.
Evaluating Decision Reliability and Quality
- Review of JevBench and other decision-specific benchmarks.
- Distinction between raw accuracy and probability calibration.
- Application of Brier scores to assess confidence reliability.
- Identifying tasks where decision models are inappropriate tools.
Requirements
- Fundamental understanding of large language models and prompt-driven application development.
- Proficiency in using command-line interfaces and Python scripting.
- No prior experience in machine learning model training is necessary.
Target Audience
- AI engineers assessing the viability of decision models for autonomous agent workflows.
- Software developers seeking to optimize LLM interactions by implementing bounded decision logic.
- Solution and platform architects developing robust agent-based system architectures.
- Technical product managers evaluating available decision model technologies.
