Selected work that connects silicon photonics, quantum computing, and artificial intelligence.
Demonstrated a photonic processor running advanced AI models (ResNet, BERT, Atari deep reinforcement learning) with near-electronic precision. Marks a substantial step towards post-transistor computing technologies.
Demonstrated a fully integrated photonic deep neural network on a single chip with forward-only training, achieving 92.5% accuracy on classification tasks with 410 ps latency. Opens path to nanosecond inference at femtojoule-scale energy efficiency.
Experimentally demonstrated quantum speed-up in reinforcement learning using an integrated nanophotonic processor. Shows systematic quantum advantage for learning agents with fast active feedback.
Introduced variational quantum unsampling, a quantum neural network approach for verification and inference of quantum circuits. Applications include quantum measurement, tomography, sensing, and ansatz validation.
Mapped quantum transport regimes using a programmable 88-beamsplitter nanophotonic processor. Over 64,400 experiments revealed environment-assisted transport and the quantum Goldilocks regime.
Demonstrated optical neural network architecture using 56 programmable Mach-Zehnder interferometers on a silicon chip. Achieved vowel recognition with potential for enhanced speed and power efficiency over electronics.
Developed scalable flip-chip integration of superconducting single-photon detectors on photonic circuits. Achieved first on-chip quantum correlation measurements with 100% device yield.
Generated and routed quantum-correlated photon pairs on silicon with 95 dB pump filtering. CMOS-compatible fabrication enables integration of sources and detectors on the same chip for large-scale quantum circuits.