A-PXM
A parallel execution model for agentic AI with an execution IR, compiler pipeline, runtime, and scheduler.
•
1 min read
Ph.D. Candidate in Electrical and Computer Engineering
PhD Electrical and Computer Engineering
2021-08-01
2026-12-31
University of Delaware
Bachelor of Engineering
2015-01-01
2020-12-31
Pontificia Universidad Javeriana
I work at the boundary between compiler and runtime co-design, heterogeneous systems, and agentic AI infrastructure.
My research focuses on execution contracts: explicit declarations of State, Dependency, and Effect that make scheduling and orchestration semantics compiler-visible rather than implicit. I use that idea to improve overlap, distributed execution, workflow validation, and system reliability.
The recurring themes across my work are:
Open-source systems and research infrastructure I actively build.
A parallel execution model for agentic AI with an execution IR, compiler pipeline, runtime, and scheduler.
A local-first system for experiment tracking, research context, and reproducible technical workflows.
Benchmark suites and harnesses for evaluating contract extraction and event-driven execution strategies.
Talk on compiling agentic AI programs into a contract-aware dataflow execution model.
Talk on extracting execution contracts from OpenMP to enable ARTS-style event-driven execution.
Talk on discovering task-dependency graphs and enabling compile-time optimization of OpenMP tasking.
I am currently interested in systems that turn implicit execution behavior into analyzable program structure.
That includes compiler-visible dependence and effect models, contract-aware agent runtimes, heterogeneous task placement, and tooling that makes systems research easier to reproduce, inspect, and extend.