Skip to main content
Back to challenges
Challenge

Model Hardware Profiler: AI SDK & OpenAI o3 Accelerator Selection Engine

Deploy an automated accelerator selection profiler using Vercel AI SDK and OpenAI o3 to evaluate hardware specifications across H100, Trainium2, and TPU v5p accelerators for LLM workloads. Build a constraint-matching router that selects optimal hardware architectures, targeting a 95% accuracy on workload cost-throughput optimization.

EngineeringHosted by Vera
Sign in free to start
Challenge brief

What you are building

The core problem, expected build, and operating context for this challenge.

Build an AI SDK-powered profiling agent using OpenAI o3 reasoning to dynamically match LLM inference workloads with optimal accelerator hardware specifications.

Delivery guide

How work is evaluated

Evaluation

Evaluates the precision of hardware recommendation profiles against known benchmark workloads.

Datasets

Challenge datasets

Public datasets, plus the private datasets you have access to.

Evaluation rubric

How submissions are scored

These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

schema_conformance

Ensures the engine output conforms to MHS accelerator recommendation spec.

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

recommendation_accuracy

Measures agreement with optimal hardware cost-throughput Pareto frontier. • target: 0.95 • range: 0-1

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

What you should walk away with

  • Implement multi-accelerator profiling logic using Vercel AI SDK with OpenAI o3 integration

  • Parse public hardware spec sheets (Nvidia H100, AWS Trainium2, TPU v5p) into standardized MHS constraint formats

  • Build dynamic routing algorithms to optimize cost per token and request latency under SLA constraints

  • Validate workload resource estimates against empirical vLLM baseline runs

Resources and assets

Reference links and supporting material

Dataset notes

Simulated hardware specifications and token profiling dataset derived from open benchmarking specs of Nvidia, AWS, and Google Cloud accelerators.

Challenge setup

Evaluates the precision of hardware recommendation profiles against known benchmark workloads.

Challenge input

JSON object with request rate, prompt length, output length, and latency budget.

AI SDK

Used for standardizing model orchestration and structured schema generation in TypeScript.

OpenAI o3

Provides deep reasoning over hardware constraints and trade-offs.

Evaluated output

JSON object containing selected accelerator, estimated latency, calculated cost, and architectural justification.

Evaluation checks
  • Ensures the engine output conforms to MHS accelerator recommendation spec.
  • Measures agreement with optimal hardware cost-throughput Pareto frontier. • target: 0.95 • range: 0-1
Configured evaluation evidence
  • Benchmark: MLPerf Inference 4.0
  • Recommendation Accuracy target: 0.95
  • 1 public reference case
Configured runtime
  • JavaScript execution harness
  • javascript sandbox (unavailable on Versalist)

Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.

View technical recipe

Configured tools

Action Space
  • OpenAI · Required
  • OpenAI o3 · Optional
Policy Serving
  • o3 · Optional

Evaluation contract

  • schema_conformance · Weight 1
  • recommendation_accuracy · Weight 1

Metrics

  • recommendation_accuracy · Target 0.95

Benchmarks

  • MLPerf Inference 4.0
  • general

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON object with request rate, prompt length, output length, and latency budget.

    Output: JSON object containing selected accelerator, estimated latency, calculated cost, and architectural justification.

    Execution harness: JavaScript

Evaluation tools

  • AI SDK · Agent Framework
  • OpenAI o3 · Reasoning LLM

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Model Hardware Profiler: AI SDK & OpenAI o3 Accelerator Selection Engine