ENGINEERING / OPEN SOURCE
C++ tools for scientific computing and performance-sensitive systems. Each project explores a specific computational problem, with source code and implementation details on GitHub.
01 / SYSTEMS & MACHINE LEARNING
Low-latency tree inference
Single-row inference for large ensembles of shallow decision trees. QLeaf investigates how memory layout, CPU affinity, persistent GPU kernels, and host–device coordination affect end-to-end latency.
A focused benchmark and implementation toolkit for latency-sensitive inference. The current interface returns a scalar raw margin; benchmark results depend on the model, hardware, and execution policy.
02 / COMPUTATIONAL PHYSICS
Exact quantum many-body simulation
A toolbox for constructing constrained Hilbert spaces, quantum operators, and Hamiltonians. Sparse-matrix and Krylov methods support ground-state calculations, real-time evolution, and Lindblad dynamics.
Designed for flexible small-to-medium quantum systems, including fermionic and bosonic degrees of freedom. Used in research on quantum-light control and photodoped superconductivity.
03 / DATA STRUCTURES
One object store, multiple views
An object store with stable addresses and compile-time index definitions. Objects can participate in optional or partitioned views while storage and object lifetime remain centrally managed.
Built for bounded working sets and repeated queries, with explicit indexing control and an optional fixed-size pool. The design addresses consistency across indices without duplicating the stored objects.
04 / LEARNING ALGORITHMS
Composable reinforcement learning
A lightweight reinforcement learning framework that separates agents from their value models. SARSA and Q-learning agents work with tabular storage or Torch-based function approximators.
Template-based interfaces keep the interaction loop small: observe a state, select an action, and collect a reward. Core components are header-only; neural models use LibTorch.
05 / COLLABORATIVE RESEARCH SOFTWARE
Nonequilibrium systems simulation
A shared research library for nonequilibrium Green’s functions and interacting quantum systems. Version 2.0 adds memory-truncated time propagation and steady-state calculations.
I contributed to the development of NESSi 2.0 through my research on nonequilibrium many-body methods. The package supports accurate numerical integration and distributed-memory calculations for quantum dynamics.
RESEARCH ↔ SOFTWARE
QuDrip supports research on light-controlled quantum matter and photodoped superconductivity. The publication record also includes the collaborative NESSi 2.0 simulation package.