|
2022 - today
|
|
SRI International,
Artificial Intelligence Center - Senior Computer Scientist:
|
|
PARADIGM:
PARACHUTE, a Prolog-based framework for abductive trace
completion and projection. It tracks and guides medical procedures in
real time, under noisy and incomplete observations, as part of the
AMIRA task guidance system in the ARPA-H PARADIGM program.
KEPS 2026 paper.
|
|
COLLEAGUE:
a multi-agent Hierarchical Task Network (HTN) planner for
SRI's collaborative robot framework, with partially ordered plans,
reasoning about pre- and postconditions, and agent-aware task
decomposition.
|
|
KMASS:
research on large language models for knowledge management in
the oil and gas domain, and Concept Vectors, an interpretable topic
and user model built from multi-label concept classifiers.
Paper on the deployed system.
|
|
COBRA:
a test and evaluation harness for a federated mission-plan repair system.
|
|
Pathway Tools:
new search facilities, and a six-fold reduction of the
compilation time of the Lisp code base. I also refactored Franz's CLIM,
the GUI toolkit Pathway Tools is built on, to the current
foreign-function interface of Allegro Common Lisp (def-foreign-call
and def-foreign-type in place of defforeign and def-c-type), tested on
64-bit Linux and Apple Silicon. This is now part of
Franz's CLIM repository.
|
|
|
2017 - 2022
|
|
SRI International,
Artificial Intelligence Center - Advanced Computer Scientist:
|
|
Procedural knowledge capture from video:
an ontology-driven multimedia integration architecture that
extracts procedural knowledge from how-to videos, using OWL and SPARQL.
This work contributed to a US patent.
|
|
MEDIC and CHATPAL:
two virtual personal assistants built with OntoVPA - one for
Army combat medics, with a
domain-specific language for authoring medical dialogue workflows,
and a personalized companion for elderly users that learns
facts about its user in conversation.
|
|
Facade classification:
deep neural networks that classify building facades by
architectural style, trained in PyTorch and deployed as a REST service.
|
|
|
2010 - 2017
|
|
SRI International,
Artificial Intelligence Center - Computer Scientist:
|
|
OntoVPA:
an ontology-based dialogue management system for virtual
personal assistants. It uses OWL and RDF for the discourse
representation and SPARQL rules for the dialogue dynamics, and was
deployed for three commercial customers. See the
IWSDS 2017 paper
and the
Common Lisp framework presented at ELS 2022,
with a
video of the talk.
|
|
AURA / Project Halo:
I maintained, modernized and scaled the knowledge
representation and reasoning backend of the AURA intelligent biology
textbook:
standard knowledge base exports
(OWL2, TPTP, ASP, SILK, first-order logic) of the knowledge base
KB_Bio_101,
a new ABox-like representation language, and a new
high-performance query language. I also worked on question answering
algorithms.
|
|
Sunflower:
refactoring and optimization of the ontology management layer
of the Sunflower reasoner for Flora-2, and a graphical ontology
editor.
|
|