Skip to main content
Back to challenges
Challenge

Multi-Agent Inference Engine Auto-Tuner with AutoGen and vLLM

Automate continuous batching engine parameter tuning for specialized agent workloads. Build a multi-agent conversational optimization system in AutoGen where specialized agents collaborate to benchmark and adjust vLLM engine flags like max_num_seqs and block_size to maximize throughput.

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 AutoGen multi-agent system that iteratively configures, benchmarks, and tunes vLLM inference engine hyper-parameters to find optimal throughput configurations.

Delivery guide

How work is evaluated

Evaluation

Evaluates auto-tuner optimization efficiency in discovering high-throughput engine configurations.

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

Throughput Improvement Test

Ensure tuned config outperforms baseline by at least 15%.

Binary check

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

Dimension 2

Throughput Optimization Factor

Ratio of tuned throughput vs theoretical optimal throughput. • target: 0.92 • 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

  • Construct AutoGen conversational agent loops between Tuner, Tester, and Analyzer agents.

  • Iteratively search vLLM configuration parameter spaces (`gpu_memory_utilization`, `block_size`).

  • Identify high-throughput engine configurations for target agent benchmark tasks.

Resources and assets

Reference links and supporting material

Dataset notes

vLLM performance benchmark log dataset containing varying parameters and throughput outcomes.

Challenge setup

Evaluates auto-tuner optimization efficiency in discovering high-throughput engine configurations.

Challenge input

JSON target_workload_type, baseline_throughput_tps

AutoGen

Multi-agent conversational interaction framework.

Evaluated output

JSON optimal_config, achieve_throughput_tps, throughput_gain_percent

Evaluation checks
  • Ensure tuned config outperforms baseline by at least 15%.
  • Ratio of tuned throughput vs theoretical optimal throughput. • target: 0.92 • range: 0-1
Configured evaluation evidence
  • Benchmark: Inference Engine Hyper-parameter Search Standard
  • Throughput Optimization Factor target: 0.92
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python 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
  • AutoGen · Required
  • Anthropic · Optional
  • Microsoft AutoGen · Optional

Evaluation contract

  • Throughput Improvement Test · Weight 1
  • Throughput Optimization Factor · Weight 1

Metrics

  • Throughput Optimization Factor · Target 0.92

Benchmarks

  • Inference Engine Hyper-parameter Search Standard
  • general

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON target_workload_type, baseline_throughput_tps

    Output: JSON optimal_config, achieve_throughput_tps, throughput_gain_percent

    Execution harness: Python

Evaluation tools

  • AutoGen · Agent Framework

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Multi-Agent Inference Engine Auto-Tuner with AutoGen and vLLM