Apache Mahout — Hackathon at Community over Code Glasgow 2026

Apache Community over Code 2026 — Glasgow, UK, October 11–14

Register now | Event website | Hackathon overview

Apache Mahout — Hackathon

Coordinator

What We’re Working On

Apache Mahout is a quantum computing project focused on two major components:

Installing Qumat and QDP

Install Qumat from PyPI:

pip install qumat

Install with QDP (Quantum Data Plane) support:

pip install qumat[qdp]

from qumat import QuMat

qumat = QuMat({
    "backend_name": "qiskit",
    "backend_options": {"simulator_type": "aer_simulator"},
})
qumat.create_empty_circuit(num_qubits=2)
qumat.apply_hadamard_gate(0)

Change backend_name to cirq or amazon_braket and the same code runs elsewhere. Pure Python — it needs no special hardware.

QDP is where the systems programming lives: memory layout, kernel performance, streaming pipelines, unsafe audits. It needs Linux and an NVIDIA GPU to run.

Useful background that can help:

Pre-installs that can save time in the hackathon

Read this before installing anything.

QDP requires Linux + an NVIDIA GPU + the CUDA toolkit. We might not be able to run those tasks as many of us is going to be with a mac only.

Setup

Do this before you arrive if you can. Half an hour at home saves you an hour in the room.

Everyone: base setup

Requirements: Python 3.10, 3.11, or 3.12 (not 3.13 — see requires-python in pyproject.toml), git, and uv.

# 1. Install uv if you don't have it
pip install uv

# 2. Clone and install
git clone https://github.com/apache/mahout.git
cd mahout
uv sync --group dev

# 3. Install the pre-commit hooks (formatting and lint run automatically)
pre-commit install

Verify it worked:

make test_python
Track What Hardware Languages
A Qumat — circuits and quantum ML Any laptop Python
B Docs, examples, website Any laptop Markdown, JS
C QDP internals — refactors and review Any laptop (compiles, cannot run) Rust
D QDP GPU work Linux + NVIDIA GPU + CUDA Rust, CUDA, Python

If you have a Mac or Windows laptop: Tracks A, B, and C. All three have real, mergeable work in them.

During the hackathon we will focus on:

Hackathon Tracks

Track A — Qumat · Python · any laptop

Task Level
A1 Backend parity sweep. Pick a gate or feature and check it behaves identically on Qiskit, Cirq, and Braket. Fix or document what differs. Self-contained, and a genuinely good first PR. [new]
A2 #749 — Parameter optimization utilities for parameterized quantum circuits [some]
A3 #1409 — Qumat QMLTrainer, a training loop for quantum ML models [some]
A4 #456 — Quantum kernel methods research spike, and #458 on downstream kernel calculations. The mathematical end. [deep]

Track B — Docs, examples, website · any laptop

Task Level
B1 Break our quickstart. Follow Getting Started on your own machine and file an issue for every step that fails or confuses you. This sounds trivial; it is not. You can only see those problems once, and we cannot see them at all any more. No code required, and one of the most valuable things you can do. [new]
B2 #1432 — Document Preprocessor in the qumat_qdp public API [new]
B3 #723 — NumPy input documentation and more tests [new]
B4 Add a worked example. examples/ is thin. A clear, well-commented notebook is very welcome. [new]
B5 #1359 — Document the API docs contributor workflow [some]
B6 #1308 — Replace website/scripts/sync-docs.js with Docusaurus-native content loading. JavaScript; good for web developers. [some]

To preview the website locally: cd website && npm run start.

Track C — QDP internals · Rust · any laptop

Task Level
C0 Verify the no-GPU build. Run cargo check -p qdp-core on a machine with no CUDA toolkit, confirm whether the stub fallback works, and document the result in qdp/DEVELOPMENT.md. Small, and it unblocks every future contributor without a GPU. [new]
C1 #1331 — Many QDP tests construct a QdpEngine or use device="cuda" with no GPU guard. Our pick of the list: not having a GPU is exactly what lets you verify the fix, and you make the project easier for the next person in your position. [new]
C2 #1436 — StreamingProducer: use VecDeque for O(1) buffer advance [some]
C3 #1433 — KernelElem trait to deduplicate the f32/f64 stubs [some]
C4 #1434 — Unify PipelineIterator constructors with a Source enum · #1435 — Unify batch producers via AdapterBackedProducer [some]
C5 #787 — NumPy file size check to prevent OOM [some]
C6 #1009 — Audit and reduce unsafe usage in QDP. Ongoing; reviewing it teaches you the codebase fast. [deep]

Resources

Who Should Join?

We welcome:

Prior experience with quantum computing is not required. Most tasks involve standard Python, Rust, JavaScript, documentation, and software engineering skills.

Submitting Your Pull Request

Full detail is in CONTRIBUTING.md and the PR policy. The short version:

git checkout -b your-feature-name
# ... make your changes ...
git add .
git commit -m "Description of your changes"   # pre-commit hooks run here
git push origin your-feature-name

Then open the PR against apache/mahout and fill in the template.

What reviewers will look for:

Tests live in testing/qumat/, testing/qdp/, and testing/utils/.

Your PR does not have to be finished by 18:40. Nobody expects a completed feature in one afternoon. An open draft PR with a clear description, or even a well-written issue documenting what you found, is a real contribution. Plenty of people carry on after the conference.

Getting Help

Turning up with a laptop and some curiosity is enough. See you in Glasgow.