"NAUTILUS team on AI Infra Summit Hackathon Hackathon"
Team Idea
GPUPilot — AI-Powered GPU Infrastructure Optimizer GPUPilot is an autonomous AI infrastructure engineer designed to monitor, analyze, and optimize AI workloads running on GPU-based systems. It continuously tracks GPU utilization, memory usage, temperature, power consumption, inference latency, throughput, and other system metrics to identify performance bottlenecks. Using AI agents, GPUPilot analyzes the collected telemetry, determines the root cause of inefficient workloads, and recommends or applies optimizations such as batch-size tuning, precision optimization, quantization, memory optimization, and workload scheduling. It then benchmarks the workload before and after optimization to measure the actual performance improvement. The system follows a continuous Observe → Diagnose → Optimize → Benchmark cycle, helping developers achieve higher AI inference performance while reducing infrastructure cost and resource consumption.
Team Members
.png&w=256&q=75)
Harshit Yadav
harshityadav165391527
