AI Agent Instructor

Job Type: Contract

Location: Remote

Archal Labs is an incipient startup backed by Y Combinator (S26). We are building the evaluation infrastructure for the next generation of autonomous agents. Our work focuses on creating complex, long-sequence agentic workflows that serve as the "ground truth" for training large models, or for fine-tuning existing models for difficult tasks.

The Role

We are looking for motivated, analytical, detail-oriented students. You will record high-quality trajectories of real workflows, step-by-step examples of how an ideal AI agent should execute a complex task across multiple applications. Each task is remote and able to be done on your own time.

In particular, our early hires will work closely with the founders to improve the software platform and trace collection process. 

We are currently only hiring students from the following states: FL, TX, AZ, UT, SD, WY, KY, MS, AL, MI, NC, IA, KS, SC, ND, WV, TN, MO.

What You’ll Do:

Record Expert Traces: Manually execute complex multi-step workflows (e.g., "Research X, find data in Y, and summarize in Z") to create training data. This might look like comparing prices for a particular desk across various vendors, navigating using Google maps, or tasks on e-commerce sites.

If you have experience in proprietary, industry specific software, please mention so.

Edge Case Discovery: Identify where current SOTA models fail and document these failure modes.

Model Evaluation: Use our proprietary harness to test agent performance against your recorded benchmarks.

Who You Are:

Tech-Fluent: You are comfortable learning new software quickly and understand the basics of how LLMs/Agents work.

For any hiring related questions, feel free to contact us at hiring@archal.ai