Quantum Photonics & AI Group, MIT

The Quantum Photonics & AI Group at MIT works at the intersection of quantum photonics and artificial intelligence. We develop photonic hardware for computation, communication, and sensing at the quantum limit, alongside the AI methods that design, program, and verify such systems. Our research spans optical neural networks and photonic processors, computational design and nanofabrication of foundry-ready devices, quantum networks and diamond-based quantum sensors, and formal verification of scientific reasoning in Lean 4.

Research

Physical AI — Photonic neural networks · RF-photonic processors · Optical tensor cores · Neuromorphic computing

Devices & Computational Design — Photonic device simulation · PDK design · Nanofabrication · Programmable photonic circuits

Quantum Networks & Computing — Quantum repeaters · Entanglement distribution · Modular quantum computing · Quantum internet

Quantum Sensing — Diamond NV centers · Spin-photon interfaces · Quantum magnetometry · Color center engineering

Formal Verification in Science — Lean 4 theorem proving · Automated claim verification · AI-augmented discovery · Knowledge graph proofs

PixCell

Our first public release: a representation-first image-to-code curriculum for photonics. PixCell pairs high-visibility photonic component images and physical footprint context with verified, primitive-only GDSFactory programs — from pixels to foundry-ready parametric cells.

from datasets import load_dataset

core = load_dataset("qpaig-mit/pixcell", "core-v1", revision="core-v1", split="train")
depth = load_dataset("qpaig-mit/pixcell", "depth-v1", revision="depth-v1")

No custom loader or trust_remote_code required.

PixCell accompanies our paper From Pixels to PCells: A Neurosymbolic Approach to Photonic Component Creation (preprint forthcoming).