Applied AI Engineer - Build Next-Gen Agent Systems

Perplexity; San Francisco (CA); On-site; USD 220,000 - 405,000; Full time

Job summary

Perplexity is looking for an Applied AI Engineer to design, build, and iterate on cutting-edge agents powering our core Perplexity Computer experience. You will collaborate with engineers, PMs, data scientists, and designers to advance ML models, run experiments, and deploy production-ready systems that impact millions of users.

You will own the full AI lifecycle from research to production, applying state-of-the-art techniques, running rigorous experiments, and shipping robust, scalable

Qualifications

• 5+ years experience building and shipping AI products.
• Strong Python and software engineering skills.
• End-to-end experience with data analysis, modeling, evaluation, and deployment of AI systems.
• Excellent collaboration in cross-functional teams.
• Advanced degree or equivalent practical experience.

Responsibilities

• Apply state-of-the-art ML and LLM techniques across personalization, context summarization, retrieval and ranking.
• Build auto research harness for offline and online experiments with metrics.
• Own the model lifecycle from research to production, including A/B tests and iterative improvement.
• Collaborate with engineers, PMs, data scientists, and designers to drive product improvements.
• Stay at the forefront of ML/AI by evaluating new research and algorithms.

Skills

Python

Production quality code

Agentic tooling

ML lifecycle

Cross-functional collaboration

Education

BS/MS/PhD in Computer Science or Engineering

Tools

LLM tooling

Job description

Perplexity is looking for an Applied AI Engineer to design, build, and iterate on cutting-edge agents powering our core Perplexity Computer experience. You will collaborate with engineers, PMs, data scientists, and designers to advance ML models, run experiments, and deploy production-ready systems that impact millions of users.

You will own the full AI lifecycle from research to production, applying state-of-the-art techniques, running rigorous experiments, and shipping robust, scalable