PETER C DILLINGER, Ph.D. peterd@gatech.edu http://www.peterd.org

2110 N 89th St Seattle WA 98103 404-509-4879

.

Overview	My work in software has focused on data structures and algorithms for applications including software tools that support software engineering and data storage and retrieval. My core strength is seeing many paths to “success,” so I'm often the person consulted when others think they're stuck.
Highlights	♦ Co-inventor of ribbon filter, a near-ideal Bloom filter replacement for memory efficiency, now used in RocksDB. Best paper award at Symposium on Experimental Algorithms 2022 ♦ Key developer and project lead in adapting and extending the legendary Coverity static analysis engine, for C/C++ bug finding, to find bugs with high accuracy in Java, C#, JavaScript, PHP, Python, Ruby, Swift, and VB. https://www.synopsys.com/blogs/software-security/author/pdillinger/ ♦ Inventor of a fast, scalable, and accurate method of detecting mistyped identifiers in dynamic languages such as JavaScript, PHP, Python, and Ruby without use of a natural language dictionary. Patent US10216611B2. Coverity feature: https://stackoverflow.com/a/34796105 ♦ Did the impossible with git: on wanting to “copy with history” as part of a refactoring, I quickly developed a way to do it despite the consensus wisdom. https://stackoverflow.com/a/44036771 ♦ Did the impossible with Bloom filters: made the data structure simultaneously fast and accurate with a simple hashing technique, now used in tools including LevelDB and RocksDB. https://en.wikipedia.org/wiki/Bloom_filter (Search "Dillinger")
Work	Facebook / Meta, October 2018 to present Software engineer, mostly on RocksDB (600+ public pull requests authored and closed) ♦ TODO: Coverity, August 2009 to October 2017, acquired by Synopsys in 2014 Principle software engineer, tech lead, and manager for static and dynamic program analysis projects. ♦ New code checkers (two patents): Coverity software automatically finds coding mistakes that matter to programmers, by going much deeper and by being more clever and more cautious than linters. I conceived, designed, implemented, and/or oversaw many successful checkers for new bug patterns. The mistyped identifier checker was a solo side project while a manager. ♦ Full support of new languages: I was a key contributor (Java and JavaScript) or project lead (C#) in adapting the Coverity engine to support new programming languages, successfully meeting goals for defect density and false positive rates to claim full support for these languages. ♦ Fast support of new languages: When highest leadership called for rapid expansion of language support, I architected, negotiated, and delivered on a strategy for rapid support of new languages (PHP, Python, Ruby, Swift, and VB) that compromised defect depth and variety while providing defect quality and credibility, and left a good technical path to full language support. Ultimately, middle management knew the project was all but doomed without a strategy meeting my approval, and after my compromise strategy came in on scope and on time, no one was saying “if only ...”. → This job included tons of reading and understanding random open-source code, in all languages listed above, to find false positive defect reports on real code before customers do. → Most development was in C++ on Linux, plus some in Java, some on Windows, etc.
Education	Doctor of Philosophy in Computer Science · Northeastern University, 2007 - 2010; transfer with advisor from Georgia Tech, 2003 - 2007 ♦ Adaptive Approximate State Storage (thesis research): Explicit-state model checkers use approximate (probabilistic) state storage techniques such as Bloom filters to reduce memory requirements by orders of magnitude while maintaining very high defect detection ability. My Bloom filter improvements (in highlights above) landed me an internship at JPL where I integrated them into the award-winning model checker Spin. But the problem remained that the highest accuracy for available memory was only possible with unrealistic prior knowledge of the reachable state space size. I solved this problem with a much more advanced system for fast, approximate state storage that adapts on-the-fly to an arbitrary state space size. An information theoretic argument shows that a real user of my approach with 2m bytes of memory will be at least as accurate as, and about as fast as, an oracle with m bytes of memory. ♦ ACL2s: "The ACL2 Sedan" (assistantship research, 2005 - 2009): I am the initial developer of ACL2s, an Eclipse-based development environment for the award-winning ACL2 theorem proving system. This tool sucessfully lowered barriers to entry so that freshman students could be taught to use an industrially successful automatic theorem proving system (see 'teaching' below). Master of Science in Computer Science · Georgia Tech · 2002 - 2003 Bachelor of Science in Computer Science · Georgia Tech · 1999 - 2002   3.70 / 4.00 GPA, highest honors
Newer Publications	♦ On the Privacy of Multi-Versioned Approximate Membership Check Filters. IEEE Trans. Dependable Secur. Comput. 2024. ♦ Optimal Uncoordinated Unique IDs. PODS 2023. ♦ Fast Succinct Retrieval and Approximate Membership Using Ribbon. SEA 2022. ♦ Ribbon filter: practically smaller than Bloom and Xor. Tech report, 2021. ♦ Approximate Membership Query Filters with a False Positive Free Set. Tech report, 2021. Coauthors: Stefan Walzer, Guido Tagliavini, Pedro Reviriego, Alfonso Sánchez-Macián, Peter Sanders (Erdős number of 2), Lorenz Hübschle-Schneider, Martin Farach-Colton.
Older Publications	♦ Fast, All-Purpose State Storage. SPIN Workshop, 2009. ♦ Hacking and Extending ACL2. ACL2 Workshop, 2007. ♦ ACL2s: "The ACL2 Sedan". Workshop on User Interfaces for Theorem Proving (UITP), 2006. (Another version was a demo at ICSE 2007.) ♦ Bloom Filters in Probabilistic Verification. Conference on Formal Methods in Computer-Aided Design (FMCAD), 2004. ♦ Fast and Accurate Bitstate Verification for SPIN. SPIN Workshop, 2004. Coauthors: Panagiotis Manolios, Matt Kaufmann (Erdős number of 2), J Moore, and Daron Vroon.
Internships	NASA Ames Research Center · Robust Software Engineering group · Summer 2006 ♦ Revamped state storage for Java PathFinder, among other contributions to the tool. Worked under Willem Visser, Corina Pasareanu, and Peter Mehlitz NASA/Caltech Jet Propulsion Laboratory · Laboratory for Reliable Software · Summer 2004 ♦ Enhancements to state storage for Spin model checker, under Gerard Holzmann and Rajeev Joshi
Other Information	U.S. Citizen