San Francisco seed-stage startup Relari is building testing and evaluation infrastructure for autonomous systems, spanning software agents and physical robotics hardware. Founded in 2023 and backed by Y Combinator's W24 batch and E14 Fund, the company originally established itself through open-source software tools like continuous-eval and Agent Contract. Recent hiring reveals a direct expansion into robotic manipulation, EMG sensors, and neuromuscular data collection.
The company currently lists five open positions, focusing on robotics software engineering, hardware prototyping, human biomechanics research, and data operations in San Francisco.
Autonomous Vehicle Roots and Data-Driven AI Evaluation
Relari was co-founded by CEO Yi Zhang and CTO Pasquale Antonante. Yi Zhang previously managed product teams for autonomous vehicles at Pony.ai and warehouse robotics at Dexterity.ai after completing a computer science degree at Northwestern University and an MBA at Harvard Business School. Pasquale Antonante completed a PhD at MIT centered on the reliability and safety of complex AI systems, such as self-driving cars, alongside past engineering roles at NVIDIA and Raytheon Technologies.
In an interview with Tesoro AI, Yi Zhang explained that testing generative AI applications faces challenges similar to autonomous vehicles sticking in long demo phases. While real-world data collection can be slow or cost-prohibitive, synthetic data allows developers to simulate edge cases and stress-test systems across virtual scenarios.
As reported by MarkTechPost, Relari built an open-source platform to simulate user behavior and validate complex Generative AI applications across text generation, code functionality, and retrieval tasks. Speaking at a developer meetup hosted by Dagger, Yi Zhang outlined the team's Agent Contract framework. The system verifies multi-agent workflows by setting preconditions, postconditions, and path conditions to ensure step-by-step compliance throughout long-horizon task execution.
Hardware, Biomechanics, and Robot Learning Search
Relari's open positions indicate an active move beyond software-only evaluation into hardware platforms that capture human movement data to train dexterous robot arms and hands.
The engineering team is hiring a Hardware Engineer on-site in San Francisco to design, fabricate, and iterate on physical systems, including in-house electromyography (EMG) sensors, contact-sensing hardware, and evaluation robot fleets. Working alongside physical prototyping, the Research Engineer, Human Biomechanics role focuses on signal-processing pipelines to model neuromuscular activity, muscle responses, and touch feedback for robot learning.
To connect sensor hardware with machine learning models, the team is recruiting a Robotics Software Engineer to build software systems for robot arms and dexterous hands that learn manipulation tasks from human data. In parallel, the AI Research Engineer position develops machine learning pipelines, deploying imitation learning models and foundation models onto physical robot hardware.
To support the pipeline of sensor datasets, Relari is hiring a Data Operations Lead to manage human-data collection, vendors, and sensor inventory in a hybrid or remote setup.
Open Roles at Relari
| Role | Location | Work Mode | Salary Range |
|---|---|---|---|
| Robotics Software Engineer | San Francisco, CA, US | Onsite | $125,000 – $175,000 |
| AI Research Engineer | San Francisco, CA, US | Onsite | $125,000 – $175,000 |
| Data Operations Lead | San Francisco, CA, US / Remote | Remote | $100,000 – $150,000 |
| Hardware Engineer | San Francisco, CA, US | Onsite | $125,000 – $175,000 |
| Research Engineer, Human Biomechanics | San Francisco, CA, US | Onsite | $125,000 – $175,000 |
