SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Mercor is an AI data company building infrastructure between human expertise and frontier AI models. The company operates a platform where millions of domain experts train frontier AI models, generating over $4 million in daily payments. Mercor Enterprise extends this to Fortune 500 companies, capturing expert workflows and translating them into AI agents. The company is a profitable Series C startup valued at $10 billion.
As a Software Engineer on the Agents team, you will own agentic features end-to-end, from scoping with researchers and operations partners through implementation, launch, and iteration based on customer feedback. You'll design and ship LLM agents, harnesses, and verifiers—including the tools, prompts, and policies that ensure reliability. The role spans full-stack development: building Python/FastAPI services and Temporal/Modal pipelines that orchestrate agent runs and human-in-the-loop review cycles; crafting Next.js/React interfaces where operators and experts interact with agents; and evolving data models to provide agents with structured context and audit trails.
You'll work on two potential team tracks: Automation (building intelligent systems that automate operational work at scale, newly formed and CEO-facing, with emphasis on 0→1 product development) or Studio (owning Mercor's evaluation system and annotation platform for RL environments, working closely with frontier AI labs). In either case, you'll build state-of-the-art RL environments with realistic enterprise apps, simulated coworkers, and rich company data rooms. You'll define agent quality metrics, build evals, instrument traces, analyze failure modes, and iterate on prompts, tools, and guardrails to raise the bar for reliability, cost, latency, and UX. Cross-functional collaboration with Product, Design, Research, and Ops teams is central to shaping agent autonomy and decision-making flows.
Tech stack includes Python, FastAPI, Django, Pydantic (backend); Next.js, React, TypeScript, Tailwind (frontend); PostgreSQL, MySQL, Snowflake, DuckDB, Redis (data); Kubernetes, Temporal, Modal (orchestration); LangGraph, LangChain, FastMCP (agents/LLM); and Datadog, PostHog, LangSmith (observability).
