> Markdown version of [/jobs/ext/3762324-ai-system-performance-benchmaking-engineer-w2-profiles-onsite](https://www.wearedevelopers.com/jobs/ext/3762324-ai-system-performance-benchmaking-engineer-w2-profiles-onsite). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# AI System Performance & Benchmaking Engineer -- W2 Profiles -- Onsite

- **Company:** Trebecon LLC
- **Location:** Atlanta, GA, United States
- **Contract:** Contract
- **Skills:** Linux, Intrusion Detection Systems, Python (Programming Language), Linux Kernel, Online Transaction Processing, SQLite, Large Language Models, Perf (Linux), Reasoning Engines, Docker
- **Published:** October 15, 2026
- **Apply:** https://www.dice.com/job-detail/eb5e14a7-7c36-4cbf-91d1-de923684eb2c

## About the Role

* Python, Linux internals, Docker, cgroups and Linux perf tools.
 * Hardware profiling.
 * SQLite and OTLP/OpenTelemetry exporters.
 * Benchmarking LLMs or agents, including inference engines and runtime configurations.

Nice to have:

 * Xeon platform tuning.
 * Terminal-Bench.
 * Experience with AgentSysPerf itself.

## Description

* Review AgentSysPerf's architecture end to end: benchmark adapters, measurements, analyzers, storage, replay, reporting and dashboards.
 * Review state persistence, container lifecycle, concurrency and the AgentInvoker protocol implementation.
 * Integrate up to 4 new agentic benchmarks (e.g., τ-bench, BrowseComp, GAIA, BFCL). This includes configuration, CLI execution, model/runtime integration, metrics capture and results persistence.
 * Measure latency, throughput and resource use on Intel hardware.
 * Improve run variance, repeatability, fault handling and SQLite write concurrency during large sweeps.
 * Add new schema fields (policy hashes, topology IDs, route decisions, resource placement) to the storage and telemetry pipelines.