Role overview
Build and run applied AI services using open-weight language models for internal workflows such as document processing, customer support, and employee chat. The work spans application development, model evaluation, and cloud deployment, with an emphasis on measurable quality, operating cost, and response time.
Responsibilities
- Connect models to approved company data and systems through APIs, with output validation, access controls, failure handling, and appropriate fallbacks.
- Work with infrastructure and DevOps colleagues to deploy models on controlled infrastructure and improve throughput, latency, and serving cost.
- Compare internal models with existing providers using repeatable evaluations, then use findings to guide model, prompt, retrieval, and fine-tuning choices.
- Integrate models with coding assistants or agents that retrieve repository code, suggest changes, and run tests within approved permissions.
- Help teams adopt these tools through practical examples and training while retaining code review and release practices.
- Track usage, compare internal operating costs with provider costs, and assess time saved alongside review effort and defect rates.
Requirements
- At least three years of experience in software engineering, machine learning engineering, or applied AI.
- A delivered LLM application or service used by real users, with clear evidence of personal contribution and results.
- Strong Python skills, including API development, data processing, and software debugging.
- Hands‑on experience in at least one area: LLM applications, model serving, model evaluation or fine‑tuning, or coding‑agent development.
- Ability to design repeatable evaluations, investigate failures, and explain trade‑offs among quality, latency, and cost.
- Ability to communicate technical designs and document results in English.
Nice to have
- Experience serving open-weight models with tools such as vLLM or SGLang, including load testing and GPU memory optimisation.
- Fine‑tuning experience with domain data and evaluation against a separate test set.
- Experience building document‑grounded retrieval systems or repository‑integrated coding agents.
Benefits and work setup
- Remote‑first work, with core hours from 10:00 to 15:00 in your local time zone and flexibility outside that window.
- Listed benefits include 28 days of paid annual leave, equipment, bonuses, and annual company retreats.