# Moustafa AbdelBaky

> Co-Founder and CTO at Disarray and Research Associate at UC Berkeley's SkyLab. Researcher of autonomous orchestration of complex systems under uncertainty, with two decades of work spanning high-performance computing, edge-native ML, mission-critical autonomous aviation at NASA, and compound AI systems for machine learning engineering.

Moustafa earned his Ph.D. from Rutgers under Manish Parashar, where he was a three-time IBM Ph.D. Fellow and won the IEEE IPDPS Ph.D. Forum Award and the Rutgers ECE Academic Achievement Award. His dissertation was nominanted for the ACM Doctoral Dissertation Award. He was a postdoctoral researcher at UC Berkeley's RISELab with David E. Culler, then served as a Technical Director at NASA Ames Research Center, where he received NASA's Early Career Achievement Medal for pioneering secure, real-time AI/ML systems advancing space missions and mixed-autonomy air traffic management. He now co-leads Disarray and continues academic research at UC Berkeley's SkyLab.

A single research question runs through his career: how can autonomous systems reliably orchestrate complex, heterogeneous components under incomplete information and dynamic conditions? His work has approached this question across four domains:

- **High-performance computing and distributed systems** (Rutgers / IBM Research): Doctoral work on CometCloud, a constraint-based declarative framework for federating supercomputers, clouds, grids, and clusters. Enabled scientific workflows across 100,000+ processors and powered applications from cancer detection at Rutgers Cancer Institute (months to minutes) to wildfire prediction at UCSD and oil reservoir simulation at UT Austin. Recognized with an ACM Doctoral Dissertation Award nomination, IEEE SCALE 2011, and ACM UCC 2015. Multi-cloud container orchestration work with IBM Research was featured in Fortune.
- **Edge computing and ML systems** (UC Berkeley RISELab): XBOS-DR, a distributed operating system for smart buildings deployed across 100+ commercial buildings in California; serverless computing across the cloud–edge continuum; deep RL for HVAC and smart-vehicle charging under demand response; and WAVE, a decentralized authorization framework for IoT (USENIX Security 2019). Co-author of a widely cited ACM Computing Surveys tutorial on blockchain systems.
- **Mission-critical AI infrastructure** (NASA Ames): Designed and led the Data & Reasoning Fabric (DRF), a secure decentralized data and reasoning infrastructure for autonomous aviation, with zero-trust architectures, transitive delegation across organizations, and 4D geospatial databases. Underpinned NASA's UTM program and informed distributed architectures for the Starling CubeSat constellation. Also led AMPLIFY, NASA's internal agentic-AI platform with human-in-the-loop oversight for mixed-autonomy air traffic management.
- **Compound AI systems for ML engineering** (Disarray, current): A compound architecture orchestrating the full ML lifecycle through a multi-agent framework with inner execution and outer governance loops, a semantic knowledge graph built from heterogeneous data and source code via static program analysis, hybrid retrieval respecting graph topology and provenance, persistent memory with quality-aware curation, and multi-model routing. The system has been validated across diverse Kaggle competitions in vision, NLP, tabular, and object detection, earning 28 medals — including nine top-10 finishes and one better-than-human result — in fully autonomous runs capped at 24 hours and a single A100 GPU each.

Three themes recur across these domains: the limiting factor for autonomy is context (the system's ability to understand its environment, track provenance, and reason about dependencies it cannot directly observe); orchestration at scale requires explicit constraint specification and runtime verification, not just capable components; and systems architecture is a multiplier that makes existing components dramatically more effective.

Looking ahead, Moustafa is interested in pushing compound AI systems toward broader scientific domains — autonomous AI for scientific discovery — and in the theoretical foundations of self-improving systems: under what conditions can a compound AI system reliably improve its own components, and what architectural constraints make such improvement stable and verifiable.

## Pages

- [Home / Biography](https://moustafa.us/): Bio, current roles, career narrative, and current research focus on compound AI systems with a recursive self-improvement angle.
- [Research](https://moustafa.us/research/): The four research domains in detail (HPC, edge ML, NASA mission-critical AI, compound AI at Disarray) plus the three throughline themes and a list of academic and industry collaborators.
- [Publications](https://moustafa.us/patents/): Selected publications, awards, and patents. (URL is /patents but the page covers all three.)

## Selected publications

- Ishihara, A. K., AbdelBaky, M., and Shetye, S. (2023). Zero-Trust Architecture for Autonomous Edge Computing. AIAA SciTech 2023.
- AbdelBaky, M. et al. (2021). DRF: A Software Architecture for a Data Marketplace to Support Advanced Air Mobility. AIAA Aviation 2021 Forum.
- Kolb, J., AbdelBaky, M., Katz, R. H., and Culler, D. E. (2020). Core Concepts, Challenges, and Future Directions in Blockchain: A Centralized Tutorial. ACM Computing Surveys 53(1), 1–39.
- Andersen, M. P. et al. (2019). WAVE: A Decentralized Authorization Framework with Transitive Delegation. USENIX Security '19.
- AbdelBaky, M., Diaz-Montes, J., and Parashar, M. (2018). Software-Defined Environments for Science and Engineering. International Journal of High Performance Computing Applications 32(1), 104–122.
- Full list: [Google Scholar](https://scholar.google.com/citations?user=DlBr9YEAAAAJ)

## Patents

- Constructing a Knowledge Graph Based on Data and Source Code and Searching the Graph (US 19/379,488) — foundation for Disarray's context graph.
- Querying and Updating a Knowledge Graph Constructed Based on Data and Source Code (US 19/379,491) — companion to the above.
- Idle Datacenter Resource Donation (US 05458-P0003A).
- Periodic Evaluation and Telerehabilitation Systems and Methods (US 12117991).

## Awards

- NASA Early Career Achievement Medal (2025)
- ACM Doctoral Dissertation Award Nomination (2017)
- IEEE IPDPS Ph.D. Forum Award (2017)
- Rutgers ECE Academic Achievement Award (2017)
- ACM UCC Cloud Challenge Award (2015)
- IEEE/ACM CCGrid Scale Challenge Award (2011)
- IBM Ph.D. Fellowship (2012, 2013, 2014)

## Affiliations and collaborators

- **Disarray** (Co-Founder and CTO): https://disarray.ai — building AI research agents for long-horizon autonomy over heterogeneous data, with a context graph at the core. Co-founded with Doris Xin.
- **UC Berkeley SkyLab** (Research Associate): https://sky.cs.berkeley.edu
- **Past affiliations**: NASA Ames Research Center (Technical Director); UC Berkeley RISELab (postdoc with David E. Culler); Rutgers University (Ph.D. with Manish Parashar); IBM Research.
- **Selected research collaborators**: David E. Culler, Ion Stoica, Joey Gonzalez, Joseph Hellerstein, Randy Katz, Raluca Ada Popa (UC Berkeley RISELab); Manish Parashar, Kirk E. Jordan (Rutgers / IBM Research); Jiasi Chen, Abraham Ishihara, Kalmanje Krishnakumar (NASA Ames / University of Michigan).

## Profiles and contact

- LinkedIn: https://www.linkedin.com/in/moustafa-a/
- Google Scholar: https://scholar.google.com/citations?user=DlBr9YEAAAAJ
- Disarray: https://disarray.ai