About this role
This role is ideal for a seasoned Software Engineer who thrives at the intersection of deep engineering and frontier AI research. The team at Turing is focused on accelerating superintelligence by creating the foundational data, environments, and tools that frontier AI labs use to improve model capabilities. This position is central to this mission, as the engineer will build the core machinery—including task-authoring tools, data pipelines, and evaluation harnesses—that researchers rely on daily. You will be operating at the critical juncture where research findings meet real-world deployment signals, ensuring correctness and reproducibility in systems that directly shape how frontier models learn to reason about complex tasks.
The successful candidate will own production systems from design through to operation, acting as a key liaison between researchers and engineering efforts. This involves building robust backend services and data pipelines, while also developing agentic workflows and safe, sandboxed execution environments for model-generated code. This role exists to ensure the integrity and reliability of the data and evaluation processes that drive model improvement. It is suited for someone who values rigorous correctness, understands the challenges of non-deterministic systems, and enjoys driving reliability improvements through strong technical documentation and code review.
What you will do
- Design, implement, and manage production systems end-to-end, covering design, rollout, and operational management.
- Develop backend services and data pipelines responsible for generating, validating, scoring, and versioning datasets reproducibly from configuration.
- Create high-performance internal interfaces, such as virtualized tables and server-side filtering UIs, capable of handling large datasets.
- Build and operate agent harnesses that include multi-step loops, tool use capabilities, retries, and structured output mechanisms.
- Construct sandboxed execution environments to ensure model-generated code runs safely and deterministically at scale.
- Trace and debug complex agent runs to pinpoint failures in tool calls, loop stalls, or grader disagreements.
- Manage infrastructure as code and CI/CD processes, debugging production issues across the entire stack to maintain reliability.
- Elevate the quality bar through rigorous code reviews, design documentation, and collaborative discussions with researchers.
What you need
- Possesses six or more years of experience building and operating production software, preferably within product companies that operate at scale.
- Requires deep proficiency in Python, specifically with production-grade frameworks like FastAPI or Django, including expertise in ORM performance, migrations, and asynchronous patterns.
- Must have strong experience with PostgreSQL, including schema design, query optimization, and indexing, alongside experience with NoSQL databases.
- Demonstrated experience in building production frontend applications using TypeScript, React, and Next.js.
- Fluency in Linux and Docker is required, along with experience using Infrastructure as Code tools like Terraform on AWS or GCP, and ownership of the associated CI/CD pipeline.
- Daily utilization of LLM coding tools such as Claude Code, Cursor, or Copilot, coupled with the critical judgment needed to assess and correct their output.
- Experience building systems on top of LLM APIs, agents, or pipelines where fragility points have been identified.
- Comfort with non-deterministic systems, allowing the ability to distinguish real regressions from noise and favor controlled experiments.
Nice to have
- Internal tools or developer platforms used daily by a technical team.
- Agent frameworks and protocols, including LangGraph, MCP, or OpenAI/Anthropic tool use.
- Sandboxing and untrusted code execution technologies like gVisor, Firecracker, or seccomp.
- Workflow orchestration tools such as Temporal, Airflow, Prefect, or Dagster beyond managed defaults.
- Exposure to LLM coding benchmarks like SWE-bench or eval frameworks.
Benefits
- Location: São Paulo, Brazil
Skills and tech stack
Questions candidates ask
What is the remote policy for this role?
The job is listed as Full-time and the location is specified as São Paulo, Brazil. Specific details regarding remote work flexibility are not detailed in the posting.
What is the expected compensation structure?
The posting does not specify the salary for this role.
What is the primary technology stack used?
The core backend stack includes Python with FastAPI or Django, PostgreSQL, and MongoDB. The frontend utilizes TypeScript, React, and Next.js, supported by Linux, Docker, Terraform, and cloud providers like AWS or GCP.
How is the team structured?
This role sits at the seam between engineering and research, working closely with researchers and quality owners to define what 'good' means.