SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Mastra is an open-source TypeScript framework for building AI agents, backed by Y Combinator and Spark Capital with $35M in total funding. The company has grown to 27k+ GitHub stars and 1M+ weekly npm downloads in under 18 months.
You'll work on features that help developers understand their agents' behavior, evaluate their performance, and improve them over time. This is a hands-on engineering role where you'll move between experiments, benchmarks, and product code, taking learnings from evaluating agents and using them to shape the framework.
Key focus areas include building agentic features into Mastra such as Trace intelligence and Agent Learning features (Autoevals + Suggestions, Rollouts). The team is increasingly running Mastra primitives like memory on industry-standard benchmarks and using results to fine-tune configuration.
You'll build tools for understanding traces, evaluating agent behavior, and turning results into improvements. You'll run experiments on Mastra's primitives, investigate where they perform well and where they fall short, and use those findings to improve their configuration and behavior. Your work will help turn that research into capabilities developers can use in their own applications.
The team is small, shipping fast, and fully remote with most members working across North American and European time zones.
Requirements:
- Shipped agents and agentic applications into production
- Experience tuning and tweaking agents to improve accuracy
- Ability to benchmark results against existing datasets or intelligently create new benchmarks
