Agent Platform Engineer (Remote)

Viktor•New York, NY

1d•Remote

About The Position

The Agent Platform Engineer will be responsible for building and operating the core infrastructure that powers the agent. This includes integrations with over 3,000 tools, persistent Linux sandboxes for each workspace, a tool gateway that translates tools into functions, and the underlying data layer. The role requires a deep understanding of distributed systems, sandboxing, OS-level isolation, and integration plumbing. The engineer will ensure the reliability and scalability of the platform, which is critical for the agent's success in handling real customer data. This is a systems-focused role where the engineer will have direct impact and ownership over a significant surface area of the company's operations.

Requirements

Shipped to production every day, with a changelog that has your name on it.
Ability to fix systems in production at 3 am because you understand the system end to end.
Designed for how things break at scale, with blast radius, failure modes, and cost in mind before shipping.
Built and operated distributed systems in production, with experience from when they broke at scale.
Knowledge of sandboxing and OS-level isolation: containers, Linux internals, running untrusted code safely.
Experience wrangling real integration plumbing: OAuth, webhooks, rate limits, and third-party APIs.
Understanding of multi-tenancy, durability, and cost per request, with a design for failure by default.
Instincts for security and blast radius, especially when the system executes model-written code.
Ability to build with AI by default.
Ability to learn fast.

Nice To Haves

Worked on infrastructure, platform, or developer-tools teams where reliability was the product.
Built sandboxing, code-execution, or multi-tenant systems before.
Contributed to open-source infrastructure.
Founded something or been an early-stage builder.

Responsibilities

Own the integrations: OAuth, webhooks, schema mapping, error handling, and keeping 3,000+ connectors working as the APIs under them drift and break.
Build the tool gateway: turn every connected API and MCP into a clean Python SDK the agent imports, with routing, auth, rate limits, and reliability.
Run the sandboxes: a persistent Linux environment per workspace, isolated, secured, autoscaled, and cheap per task.
Own the data layer: state and persistence across 25,000+ multi-tenant workspaces, durable and fast.
Build whatever needs building, as part of a small team with a large surface area.