SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Turing is the world's leading research accelerator for frontier AI labs. The STEM Horizontal team produces high-signal STEM data that frontier AI labs train on. This role builds the production machinery behind that mission: task-authoring tools, data pipelines, evaluation harnesses, sandboxed execution environments, and agentic workflows that researchers and domain experts use daily.
You will own production systems end-to-end, sitting at the seam between engineering and research. You'll be the engineer researchers pull in to debug why an evaluation is producing misleading numbers. Correctness and reproducibility are the product—a subtle bug doesn't crash anything but quietly poisons a training run and costs weeks of compute.
Key responsibilities:
- Own services end-to-end: design, implementation, rollout, and production operation
- Build backend services and data pipelines that generate, validate, score, and version datasets reproducibly from a commit and config
- Ship data-dense internal interfaces: virtualised tables, server-side filtering, review and annotation UIs that stay fast at tens of thousands of rows
- Build and operate agent harnesses with multi-step loops, tool use, retries, and structured output running against real repos and test suites
- Build sandboxed execution environments where model-generated code runs safely and deterministically at volume, with RL environments and eval harnesses on top
- Trace and debug agent runs end-to-end: where loops stalled, which tool calls failed, why graders disagreed with humans
- Own infrastructure as code and CI/CD; debug production issues across the stack and drive reliability work
- Raise the bar through code review, design docs, and innovation, partnering with researchers on what "good" means
Tech stack: Python, FastAPI, Django/Django Ninja, PostgreSQL, MongoDB, REST/OpenAPI (backend); TypeScript, React, Next.js with virtualised tables and server-side filtering (frontend); Linux, Docker, Terraform, CI/CD, AWS or GCP (platform).
REQUIREMENTS:
- 6+ years building and operating production software, ideally at product companies with real scale
- Deep Python with production FastAPI and/or Django experience: ORM performance, migrations, async patterns
- Strong PostgreSQL (schema design, query optimisation, indexing) and NoSQL experience
- Production frontend experience with TypeScript, React, and Next.js
- Fluent on Linux and Docker; Terraform or comparable IaC on AWS or GCP, with ownership of CI/CD
- Daily use of LLM coding tools (Claude Code, Cursor, Copilot) as part of your shipping workflow, with critical judgement of their output. This is a must-have for an AI-forward team
- Built something on top of LLM APIs that others depended on (agent, pipeline, eval) and know where it was fragile
- Comfort with non-deterministic systems: can distinguish real regressions from noise and reach for controlled experiments over hunches
- Advanced Git and real testing discipline: unit, integration, negative and edge cases
- Clear written communication; comfortable working independently and comfortable saying when a spec is wrong
Nice to have: internal tools or developer platforms used daily by technical teams; agent frameworks and protocols (LangGraph, MCP, OpenAI/Anthropic tool use); sandboxing and untrusted code execution (gVisor, Firecracker, seccomp); workflow orchestration (Temporal, Airflow, Prefect, Dagster) and Kubernetes beyond managed defaults; LLM coding benchmarks (SWE-bench, Terminal-Bench), eval frameworks, or exposure to RLHF/RLVR.
Preferred background: Bachelor's or Master's degree with 6+ years in Computer Science, Software Engineering, Data Science, Machine Learning, AI, or a programming-heavy IT field. Demonstrated work outcomes are weighted most heavily; a strong track record without a matching degree is acceptable.
