Inspiration
Small manufacturers and student builders often need optimization tools that can run locally on modest hardware without a heavyweight cloud stack. We wanted to turn manufacturing scheduling into a reproducible Arm-friendly workflow that is easy to benchmark, explain, and extend.
Track fit
This submission is best aligned with the Cloud AI track because it focuses on Arm64-friendly developer workflows, reproducible benchmarking, and optimization-oriented execution that can be validated on Arm-powered local devices or Arm64 cloud environments.
What it does
Q-Flow Edge is a lightweight manufacturing scheduling toolkit that generates synthetic scheduling scenarios, compares classical baselines against a quantum-inspired QUBO solver, and produces Markdown reports, CSV artifacts, and SVG figures. It is designed to make bottlenecks, trade-offs, and optimization results easy to inspect on an Arm64 workflow.
How we built it
The project is built with a Python standard-library-first pipeline so it stays portable and simple to run. The current repo includes a benchmark harness, a quantum-inspired annealing solver, classical baselines, recovery simulation, and evaluator-friendly submission assets. We also documented an agentic recovery architecture that can evolve into a fuller AI workflow later.
Setup and validation
Judges can clone the repository, run the project with Python 3, and regenerate the benchmark outputs with a single entry command. The repository README includes the execution flow, output artifacts, and the key documents needed to understand or validate the project.
Arm optimization angle
This project is a strong fit for Arm because it emphasizes lightweight local execution, reproducible benchmarking, and developer experience on Arm64 devices. The codebase avoids unnecessary dependencies, works well as a local toolkit, and can be benchmarked directly on Arm-powered laptops or Arm64 cloud environments.
Challenges we ran into
The hardest part was being honest about optimization claims. Instead of forcing a story that the quantum-inspired method always wins, we compared approaches on the same scenarios and preserved the trade-offs in the output reports. Another challenge was packaging the project so it could be judged quickly without losing technical depth.
What we learned
We learned that reproducibility, documentation quality, and measurement discipline matter as much as the solver itself. We also learned that an optimization project becomes much more compelling when results, interpretation, and setup instructions are all part of the same developer-friendly workflow.
Log in or sign up for Devpost to join the conversation.