active project

Odinsynth and Odinsynth-LLM

Neural and LLM-assisted program synthesis systems for generating explainable, editable extraction rules from examples.

Odinsynth explores how subject-matter experts can move from examples to editable symbolic extraction programs, combining neural search, LLM assistance, preference optimization, and human-editable rules.

Media

Odinsynth demoDemo video showing example-driven synthesis of editable information extraction rules.
Role
Research direction, system design, implementation
Period
2022-present
Audience
subject-matter experts, NLP researchers, applied scientists

Themes

Skills and Tools

PythonPyTorchTransformersScalaDocker

Links

Back to projects