Sequential Inference
Research and engineering for decisions under uncertainty.
Sequential Inference develops methods and systems for learning from data, interpreting evidence, and making decisions under uncertainty.
Our research connects statistical learning with LLMs, adaptive systems, and inference engineering. Financial markets provide the primary application setting.
The work
- ResearchStatistical learning, LLMs and agents, adaptive systems, and inference engineering.
- ModelsEducational demonstrations of foundational concepts, with their assumptions and limitations.
- SystemsEngineering for model execution, agent coordination, and controlled operation.
- PublicationsResearch and engineering work released for external review.
The standard
Our methods are assessed through defined experiments, comparisons with baselines, and explicit accounts of their limitations.
-
Test on unseen data
Separate development from evaluation and assess methods on observations excluded from their design.
-
Evaluate quality and efficiency
Measure task performance alongside latency, throughput, resource use, and computational cost.
-
Make limitations explicit
State assumptions, known weaknesses, and conditions that may invalidate a result.
-
Control changes
Evaluate proposed updates before adoption and preserve the records needed to investigate their effects.
Contact