Member of Technical Staff, Enterprise Platform

Mercor

1d•$130,000 - $500,000•Onsite

About The Position

The Mercor Enterprise Platform is what enterprise customers and our Applied AI engineers use to build and run complex agents. It is both deployed into customer environments and hosted by us. It is also the central platform for every internal agent at Mercor, scaling to several thousand agent runs per day, and still growing. You will work across the full stack: enterprise-grade security and the routing plane, isolated execution, and the intelligence plane that turns production runs into improvements. This is a pure platform and infrastructure engineering role, though you will work closely with the Applied AI engineers embedded with our customers. The quality of this platform bounds what the business can sell. Nobody has settled how agent platforms should be built, and this is one of the few places that question is being answered in production, with economically viable agents as the bar. Tech we use: Go, TypeScript, Postgres, Kubernetes, Harbor, and Terraform. Read more about how we think about this: The Three Planes of Enterprise Agent Systems

Requirements

5+ years of experience building high scale, complex software systems
You have built and operated a platform other engineers depended on, and owned the architectural calls on it rather than only the implementation.
Hands-on with several of these primitives: authn and authz, secrets, isolation, orchestration, egress, and durable state, in more than one cloud.
A drive to keep making systems faster and cheaper after they already work.
Demonstrated ability to work independently on ambiguous, loosely specified problems.

Nice To Haves

experience with LLM applications, agent frameworks or harnesses, evals, or deployed software.

Responsibilities

Design, build, and operate the services the platform is made of: authn and authz, credential brokering, routing, sandboxing, and orchestration.
Design the interfaces other engineers build agents against, and keep the paved path the easiest one to follow.
Build the event-driven backbone so that ingestion stays idempotent when services scale horizontally.
Own the deployment model itself, from our hosted offering to customer VPCs and multi-cloud, so a new deployment needs no bespoke build.
Keep tenants isolated by construction, so one run cannot reach or spend beyond its own quota.
Own latency, throughput, and cost per run as volume grows, along with the production debugging that comes with them.
Set the direction for the parts of the platform you own, not just the implementation.
Partner closely with Applied AI engineers on the hardest problems coming out of deployments. They own the customer relationship, you own the platform.