Internship: AI-based trainable decoders for qLDPC codes

Background

Quantum error correction needs a classical decoder that reads syndrome measurements and finds a correction in real time. Quantum low-density parity-check (qLDPC) codes, such as the bivariate bicycle codes, encode more logical qubits per physical qubit than surface codes. Their standard decoders, which are based on belief propagation, have trouble being fast and accurate at the same time. Recent work shows that convolutional neural network decoders that use the geometry of the code can reach lower logical error rates than belief propagation and run at high throughput on GPUs.

Project

The intern will work on our GPU neural decoder for bivariate bicycle codes under circuit-level noise. The aim is to lower inference latency without raising the logical error rate. Possible directions:

• Model compression: distilling deep models into shallow, wide ones; quantization; pruning.
• Training methods that need fewer samples, such as supervising every syndrome round instead of only the final logical outcome.
• Scaling to larger codes and lower physical error rates.
• Benchmarking on the same GPU and noise model against message-passing decoders such as relay-BP.

Requirements

• Master's or PhD student in physics, computer science, electrical engineering, or a related field.
• Good Python skills and experience with PyTorch.
• Basic knowledge of quantum error correction or stabilizer codes.

Nice to have

• GPU performance work: profiling, torch.compile, CUDA graphs, Triton.
• Experience with Stim or other QEC simulation tools.
• Experience training models at scale.

Diversity

IBM is committed to diversity in the workplace. With us you will find an open, multicultural environment. Excellent flexible working arrangements enable all genders to strike the desired balance between their professional development and their personal lives.

Interview process

After the initial screening based on the uploaded documentation, identified candidates will be contacted for a first technical discussion on their experience, background, and motivations, followed by coding interview and an AI/ML interview. Further selection steps might be added based on candidates skills and project needs.