Development, Test, and Validation Runtime Environment

Hardware and Software

Teams are encouraged to develop as much as possible on their own hardware and environments. However, to provide a fair playing field for testing and validating team submissions, contest organizers have selected the following platform upon which team submissions are evaluated:

AWS Instance m7a.2xlarge:

• 8 vCPUs: 4th Gen AMD EPYC Processors
• 32 GB RAM

AWS Vivado ML 2025.1 Developer AMI:

• Ubuntu 22.02 Operating System
• Pre-installed and licensed Vivado 2025.1

This platform should comfortably run a sequential workflow without any memory pressure issues. A larger platform with more memory and more CPU cores was avoided to enable teams to focus on optimization innovation rather than brute force parallelized exploration.

At various stages of the contest, teams may be provided with AWS credit to enable them to validate with little or no cost on the contest validation platform. Successful alpha and beta submissions may enable teams to earn additional credits. More details to follow.

LLM Access

If teams will be using LLMs for their workflow (most will), they will be required to use OpenRouter as demonstrated in the example dcp_optimizer.py agent. OpenRouter is a unified API interface for a wide range of LLMs. It provides a unified way to switch between many different LLMs using a single API. Teams may also be provided with OpenRouter credit ahead of submission checkpoints to help develop and validate their solutions. More details to follow.

Submissions must read from the environment variable OPENROUTER_API_KEY for the access key and use the OpenRouter API to access the models. No other LLM service will be supported.