@d4r5c2

David Robert Cox

Current work:

• Accelerating NP-complete computing at Integrated Reasoning (opens in a new tab) (YC S22) (opens in a new tab).
  • The biggest moonshots in YC’s S22 batch (opens in a new tab), (Kyle Wiggers (opens in a new tab), TechCrunch (opens in a new tab))

Previously:

• A memory control system introduced in iOS 15 (opens in a new tab) that leverages dynamic programming to ensure a snappy UX when changing apps under resource pressure.
• CLP(FD) (opens in a new tab) solvers for a variety of integer problems (opens in a new tab), including a handful of puzzles involving Zoombinis (opens in a new tab).
• Haskell libraries for performing inference in massive Bayesian Networks (opens in a new tab) and Markov Random Fields (opens in a new tab) using Gibbs sampling (opens in a new tab).
• Algorithms for quantifying file similarity and facilitating fast recall of data with similar information density (opens in a new tab) patterns.
• Anti-malware signatures (opens in a new tab), some (opens in a new tab) of which have been in production (opens in a new tab) since 2015 (opens in a new tab).

Publications & Preprints:

• Cox, David. "Syntactically Informed Text Compression with Recurrent Neural Networks." Mathematics Association of America, Undergraduate MathFest 61, arXiv:1608.02893 (2016) (opens in a new tab).
  • Cited by Google DeepMind (opens in a new tab): Delétang, Grégoire, et al. "Language modeling is compression." arXiv:2309.10668 (2023) (opens in a new tab).
• Cox, David. "Delta Epsilon Alpha Star: A PAC-Admissible Search Algorithm." arXiv:1608.02287 (2016) (opens in a new tab).
  • Establishes a combinatorial bound on backtracking for use in robotic search.

• GitHub @david-r-cox (opens in a new tab)
• Twitter @d4r5c2 (opens in a new tab)
• Instagram @d4r5c2 (opens in a new tab)
• Email d4r5c2@davidcox.net