Full-Time
Member of Technical Staff
Fulcrum
Agentic debugger for AI systems
Compensation Overview
$200k - $500k/yr
San Francisco, CA, USA
Hybrid
Three days on-site per week required.
 Software Engineering (2) |
|---|
| , |
- Design and build scalable infrastructure and systems for artificial intelligence agents, including containerization and performance optimization.
- Contribute to agent science research and create evaluations for agent performance and behavior.
- Develop oversight interfaces and frontend systems for coding agent monitoring and control.
- Shape product strategy and research direction in collaboration with the founding team.
Fulcrum builds tools to help developers test, audit, and improve AI agents trained with reinforcement learning. Its platform includes red-teaming agents that simulate adversarial scenarios in RL environments and an agentic debugger that surfaces the most critical data to human operators, enabling them to steer, correct, and monitor AI agents at scale. This setup uses scalable oversight interfaces and deployed auditing agents to examine entire agent ecosystems, not just individual models. Compared with typical AI testing tools, Fulcrum emphasizes end-to-end evaluation and monitoring of agent behavior and decision processes, aiming to make AI systems more reliable and transparent. The goal is to improve AI safety, reliability, and evaluation by providing practical, scalable means to audit and supervise AI agents in real time.
Company Size
N/A
Company Stage
Seed
Total Funding
$130K
Headquarters
San Francisco, California
Founded
2025
Simplify's Take
What believers are saying
- YC launched Fulcrum in September 2025, signaling strong early distribution and credibility.
- June 2026 inverse rubric optimization research shows active product and research momentum.
- Public hiring and open-source releases suggest continued founder execution and ecosystem pull.
What critics are saying
- Fulcrum sells a niche toolset into an immature market with unclear budgets.
- Open-source outputs like Quibbler invite fast imitation from bigger developer-tool vendors.
- If agent debugging stays research-only, Fulcrum becomes another YC safety lab, not a business.
What makes Fulcrum unique
- YC Summer 2025 Fulcrum targets RL-environment debugging, not generic observability.
- Its red-teaming agents inspect traces, source code, and environment interactions end-to-end.
- Quibbler and Orchestra show productized agent critique and multi-agent coordination.
Help us improve and share your feedback! Did you find this helpful?
Benefits
Company Equity
Employee Referral Bonus