AI Infrastructure Engineer - Agents & Simulation Platform
Infrastructure Engineer – Agents & Simulation Platform
About PlayerZero
PlayerZero is building a self-healing system for software that automates defect resolution and development.
We are used by engineering and support teams to:
Autonomously debug problems in the software (technical support)
Fix issues directly in the code
Prevent these problems from recurring
PlayerZero is backed by leading investors such as Foundation Capital, WndrCo, and Green Bay Ventures — and operators like Matei Zaharia, Drew Houston, Dylan Field, Guillermo Rauch, among others.
This role is a hybrid position based out of our San Francisco office. All candidates must be located in the Bay Area to be considered. Relocation assistance is not available for this position.
About the Role:
We're looking for a backend engineer who loves building the systems that make AI agents reliable, fast, and useful in production. You'll work on our real-time player sessions (interactive AI assistants), our code simulation engine, and the LLM orchestration layer that ties them together. If you enjoy designing distributed systems, tuning prompts and agents with tight feedback loops, and shipping infrastructure that developers actually trust—this role is for you.
What You'll Do:
Iterate on prompts and agent behavior: tune system prompts, tool definitions, and response parsing to improve agent reasoning, reduce hallucinations, and handle edge cases gracefully.
Design and build agent tools—the interfaces that let LLMs take action. You'll own the contract between what the model can ask for and what the system can deliver, balancing expressiveness with token efficiency.
Build retrieval and indexing pipelines that give agents the right context: chunking strategies for code, embedding pipelines, vector indexes, hybrid search (semantic + keyword), and reranking to surface what actually matters within tight token budgets.
Own the agent runtime—real-time sessions, simulation orchestration, lifecycle management, streaming, and the concurrency/cancellation mechanics that make agents reliable at scale.
Design for correctness under concurrency: distributed locks, pub/sub coordination, idempotent operations, and graceful degradation when things go wrong.
Collaborate across the stack with frontend engineers on the real-time UI, ML/AI folks on model selection and evaluation, and customers on what actually matters.
You Might Thrive in This Role If:
A builder mindset — you like owning projects end-to-end and are thoughtful about data models, performance, and long-term maintainability. Experience transitioning prototypes to production with an understanding of tradeoffs in reliability and scale.
3–6+ years of experience building backend systems—ideally with real-time, stateful, or event- driven workloads.
You've worked with LLMs in production: prompt engineering, tool/function calling, streaming responses, managing token budgets, or building evaluation harnesses.
You've built indexing or retrieval systems that give agents the right context—embeddings, vector search, or hybrid retrieval pipelines.
You're comfortable with distributed systems primitives: message queues (Kafka, Pulsar), distributed locks, pub/sub, and the failure modes that come with them.
You have attention to detail and a debugging mindset—you dig into why something failed, trace through logs and state, and don't stop at surface-level fixes.
Bonus Points:
You've tuned prompts extensively and have opinions about structured outputs, chain-of-thought, and when to use tools vs. inline reasoning.
You've built or contributed to agent frameworks, orchestration systems, or evaluation pipelines.
Hands-on with vector databases (pgvector, Pinecone, Turbopuffer) or hybrid search systems— you know when embedding similarity isn't enough and keyword matching saves the day.
You've built chunking or indexing pipelines for unstructured data (code, documents, logs) and understand the tradeoffs in chunk size, overlap, and metadata.
Experience with Kotlin or JVM-based backend services (our primary stack).
Familiarity with MongoDB, Postgres, or Redis for stateful workloads.
Hands-on with WebSockets or other real-time communication patterns.
Experience with code analysis tooling—AST parsing, Tree-sitter, or building retrieval systems over codebases.
Our Supporters
Tackling the challenges of such a vast and complex industry is an enormous undertaking. We've raised a Series A funding and are backed by leading investors who share our vision to help us on this journey.
Foundation Capital (Ashu Garg, Jaya Gupta)
WndrCo (Sujay Jaswa, ChenLi Wang)
Green Bay Ventures (Anthony Schiller, Dick Kramlich)
Matei Zaharia (Founder & CTO, Databricks)
Guillermo Rauch (CEO, Vercel)
Dylan Field (Founder & CEO, Figma)
Drew Houston (Founder & CEO, Dropbox)
Peter Bailis (CTO, Workday)
Oliver Jay (MD International, OpenAI)
John Lilly (ex CEO, Mozilla)
Bernard Kim (CEO, Match Group)