AI Software Engineer - Comet Agents

USD 200,000-300,000 per year
MIDDLE
βœ… On-site

SCRAPED

Used Tools & Technologies

Not specified

Required Skills & Competences ?

Security @ 3 Go @ 1 Playwright @ 3 TypeScript @ 1 Python @ 1 Rust @ 1 CDP @ 3

Details

Perplexity is seeking an energetic engineer to join the Comet Agents engineering team. The team consists of AI/ML, backend, and full-stack engineers who collaborate to build agentic experiences within the Comet ecosystem (https://www.perplexity.ai/comet). The role spans applied research and engineering to advance agent capabilities for a large user base.

Responsibilities

  • Drive cutting-edge AI capabilities across multiple layers of a rapidly growing product in the AI agents space.
  • Design AI agents that navigate the digital world and perform valuable units of work for users.
  • Train action and decision models that determine, based on complex multimodal states, how to accomplish user-specified objectives.
  • Develop and leverage AI models, infrastructure, and browser technologies to scale capabilities across desktop, mobile, headless cloud, and other environments.
  • Provide consistently excellent user experiences through flexible abstractions and frictionless backgrounding.
  • Develop permission architectures, payload classifiers, and secure-by-design agentic capabilities.
  • Design optimal data representations and modes of interaction between agents and environments.
  • Contribute to product reliability, code quality, AI evaluation, testing, and ongoing maintenance.
  • Collaborate with engineers, designers, product managers, and data scientists to integrate Perplexity functionality into Comet agents and vice-versa.

Requirements

  • Strong foundational familiarity with the full AI product stack.
  • Proficiency in Python (bonus points for TypeScript, Go, and/or Rust).
  • Significant experience in at least one of the following areas:
    • Context engineering and tool interfaces for frontier AI models.
    • Post-training and reinforcement learning (particularly for multimodal models).
    • Browser technologies (CDP, Playwright, extension development, etc.).
  • Experience designing or training action/decision models for multimodal state spaces.
  • Experience developing permission architectures, payload classifiers, or other security-focused agent controls.
  • Strong product intuition and a taste for user experience excellence.
  • Comfortable working on a small, fast-moving team and taking ownership of features and systems.
  • Passion for shipping products that surprise and delight users.

Benefits

  • Compensation range: $200K – $300K (offers equity).
  • U.S. full-time employees: equity, health, dental, vision, retirement, commuter and dependent care accounts, and more.
  • International full-time employees: benefits tailored to region of residence.
  • Final offers determined by experience and expertise; amounts may vary from the listed range.