SCU Hack-a-Stack - Endurance Track

About the challenge

SCU Hack-a-Stack's Endurance Track will feature 5 submission tracks. The following tracks will leverage the tech-stack or functionality of the following sponsors:

Cisco:

AI Factory Ops: Engineering the Infrastructure Behind AI Competitors will be provided with a small synthetic dataset for an AI cluster. The dataset would serve as the environment, but the challenge would be to build AI-assisted tools that support operational decision-making. To make it accessible, each subtrack functions as a constrained decision problem.

1. AI App Performance Advisor
   Given traffic, latency, error, and GPU usage data, recommend actions such as adding capacity, reducing load, rerouting traffic, or investigating errors. Example outcomes: lower latency, fewer failed requests, and better GPU usage.
2. GPU Job Placement Challenge
   Given AI jobs that need different numbers of GPUs and a set of available GPU nodes, decide where jobs should run. Example outcomes: lower wait time, fewer idle GPUs, and fewer blocked large jobs.
3. AI Job Failure Detective
   Given alerts, logs, metrics, and short runbooks, rank likely causes and recommend recovery actions such as retry, restart, move job, reduce load, or escalate. Example outcomes: better root-cause matching, faster recovery, and fewer wasted GPU hours.

SCU Ciocca Center

Adal

VoiceOS/HypeScribe

Future Unicorn

To be eligible to win a top 3 overall award, you must submit to both the Sprint track and the Endurance track.