AI Engineer, Agents & Evaluation
We’re looking for our first AI Engineer focused on agents and evaluation—a foundational hire who will shape how we build, measure, and scale intelligent systems.
The Opportunity: Design the Playbook for High-Performance AI Agents
We’re tackling one of the hardest—and most important—problems in software engineering: helping developers understand, evolve, and operate complex systems using autonomous and event-driven AI.
In this role, you’ll build the evaluation frameworks, task harnesses, and orchestration strategies that make our agents reliable, testable, and genuinely useful. Your work will not only directly improve our agents—it will create reusable benchmarks and artifacts that can inspire new approaches and push forward the broader foundation model ecosystem.
If you love designing experiments, building systems, and iterating tightly between theory and code—and you’re excited by a very 0→1, research-engineering style role—this is for you.
What You Will Do
Create Task Evaluations That Matter: Design and implement task-specific evaluations that measure and improve agent quality. Each eval should both drive concrete iteration on our agents and spark broader innovation around the task itself.
Define Tasks, Datasets, and Harnesses: Clearly specify tasks, collect and curate balanced datasets, and build robust evaluation harnesses that can be used across agents and modeling approaches. There is ample room for architectural design and systems thinking here.
Build and Use a Reusable Evaluation Framework: Develop frameworks and tools for running evaluations at scale. Use these frameworks to tune existing agents and to guide the development of new ones in our environment.
Explore Agent Orchestration Strategies: Investigate and implement orchestration patterns (tooling, routing, decomposition, multi-agent setups, etc.) that allow agents to tackle increasingly complex, multi-step, and long-horizon tasks.
Apply Post-Training Techniques: Experiment with post-training approaches (e.g., fine-tuning, preference optimization, reward shaping, distillation) to produce high-performance models tailored to specific tasks and workflows.
Run Experiments End-to-End: Design, run, and analyze experiments with rigor. Turn experimental results into clear recommendations and concrete changes to model configurations, prompts, and system design.
Collaborate Deeply Across the Stack: Work closely with founders, product, and infrastructure engineers to ensure evaluations, agents, and platform primitives all reinforce each other.
What You Will Bring
MS or Ph.D. in a relevant field (e.g., Computer Science, Machine Learning, NLP) or equivalent practical experience
Strong background in machine learning and large language models, ideally including both research and hands-on implementation
2–5 years working with LLM technology, with familiarity across:
Prompting and interaction patterns
Agent and tool orchestration strategies
Evaluation strategies for complex, open-ended tasks
Proficiency writing production-quality code, especially in Python; comfort working with TypeScript or modern web/backend stacks
Experience designing and running experiments, and interpreting results in messy, real-world settings
Self-motivated, comfortable operating in an unstructured, high-ambiguity environment
Strong communication skills and the ability to translate vague goals into concrete, testable setups
Bonus Points
Experience building agentic systems (tool-using agents, workflows, or multi-agent systems) in real products
Prior work on model evaluation frameworks, benchmarking, or reliability/robustness testing
Familiarity with modern ML tooling (training/inference stacks, experiment tracking, data pipelines)
Contributions to open-source LLM, tooling, or evaluation projects
Experience at an early-stage startup or research lab where you owned projects end-to-end
Benefits & Perks
Significant equity in an early-stage, venture-backed startup
Comprehensive Health Benefits (Medical, Dental, Vision)
Flexible PTO to ensure you have the time you need to recharge