Principal Software Engineer
Own and evolve the RL Gym platform, building isolated web environments for adversarial reinforcement learning. Architect high-performance orchestration systems, work with Firecracker/Docker/AWS, and deliver directly to AI labs. Lead a small team, move fast from concept to production.
Responsibilities
- Lead the architecture and evolution of the AgenticVerse orchestration platform for scalable RL environment provisioning.
- Design and deploy isolated web environments using Firecracker microVMs and Docker containers.
- Develop multi-site scenarios with complex interactions (drag-and-drop, file uploads, authentication) and deterministic verifiers.
- Collaborate directly with customer engineering teams to integrate environments into training pipelines, ensuring SLAs and reliability.
- Rapidly prototype new environment types and interaction patterns, iterating on feedback.
- Build internal tooling and customer-facing interfaces for environment configuration and monitoring.
Requirements
- 8+ years of software engineering experience building production systems from scratch.
- Deep expertise in Linux, Docker, Firecracker, networking, and AWS.
- Strong Python skills with experience in async/concurrent systems.
- Platform or developer tools building background.
- Full-stack capability: backend services, infrastructure-as-code, APIs, React frontend.
- Proven ability to work autonomously with minimal specification and make sound architectural decisions.
- Comfortable working directly with external customers and translating technical constraints.
- English fluency.
Nice to have
- Experience with reinforcement learning infrastructure, training pipelines, or evaluation frameworks.
- Background in security, adversarial testing, or trust & safety.
- Familiarity with browser automation, headless browsers, or web scraping.
- Experience with Kubernetes operators or custom schedulers.
- Prior work in a 0-to-1 environment (startup, innovation lab, R&D).