APPLIED AI · RESEARCH & ENGINEERING
I build AI systems, from predictive models
to tool-using agents.
My work covers NLP pipelines, retrieval, tool interfaces, and production systems.
Python tools for auditing agent datasets and checking whether recorded outcomes and costs meet a deployment policy.
Research experiments, not production ROI claims.
Requested: SCALE
Policy resultASSIST
4 of 6 required checks failed. Eight synthetic tasks; not production performance.
Inspect the source recordCo-inventor on a method that connects similar text spans in a graph, clusters them by their attributes, and maps the clusters into domain-specific groups.
Acknowledged contributor to ServiceNow’s technical review framework for AI agents.
SYSTEMS / TECHNICAL NOTE
Connecting predictive models, generative components, and agents without losing their differences.
Read noteAGENTS / ESSAY
Dynamics blindness, feedback, and the behavior of systems that act on the world.
Read essayRESEARCH / NEGATIVE RESULT
A held-out test, a negative result, and the limits of what an experiment supports.
Read findingsUCLA Extension · Instructor
ADSA 2024 · IBM Research
IBM Think · Episode 51 · Guest
IBM TechXchange 2024 · Presenter