Research grounded in mathematics, built for real systems.
AI Wranglers is built on an active research program spanning optimization, machine learning, AI safety, and AI for science. Below are the broad themes; individual projects and papers are being ported here gradually. For now, the full portfolio lives on our creator's research page.
Optimization & Scientific Computing
Provably fast optimization algorithms, high-performance computing, numerical analysis, and large-scale solvers, including work on bundle adjustment for 3D reconstruction using deflation and multigrid methods.
Machine Learning & Generative AI
Generative and diffusion models, medical AI and imaging, and extreme multilabel classification: building learning systems that hold up under real-world scale and noise.
AI Safety, Alignment & Verification
Formal verification of AI systems, and methods for building models whose behavior can be understood, tested, and trusted rather than merely observed.
AI for Science & Engineering
AI applied to scientific discovery and engineering domains, including AI-driven VLSI (chip design) and multi-agent reinforcement learning for social generalization.
Notable projects
Bundle Adjustment Solver
Fast, scalable 3D reconstruction solvers using deflation and multigrid methods.
MARL-JAX
A multi-agent reinforcement learning framework for social generalization.
FlowRead
A PDF workflow application for annotation and organization, a precursor to our training reader.
For full papers, projects, and publications
The complete research portfolio (publications, ongoing projects, and academic background) is maintained on our creator's personal page, and will be gradually ported into AI Wranglers under these broad themes.