AI Developer – Conversational & LLM Training

Job Description

About the Role

We’re hiring an Applied AI Engineer to design, build, and operate LLM- and voice-powered features in production. You’ll work primarily with hosted foundation models and real-time audio APIs rather than training models from scratch — owning prompt design, evaluation, observability, and the engineering around inference at scale.

This is a hands-on engineering role for someone who enjoys shipping AI features end-to-end:

integrating providers, designing prompts and evals, hardening real-time pipelines, and iterating from real user data.

Responsibilities

● Build and operate production features powered by hosted LLMs (OpenAI, Google Vertex / Gemini, Anthropic, or similar).

● Design and maintain real-time conversational pipelines, including STT, TTS, turn-taking, and websocket-based audio streaming.

● Own prompt engineering, versioning, tracing, and evaluation across model providers.

● Run model selection and provider migrations, balancing latency, cost, and output quality.

● Build evaluation harnesses and guardrails for safety, bias, and content quality.

● Ship behind feature flags and iterate from real user data.

● Collaborate with product, engineering, and domain experts to ground model behaviour in real requirements.

Requirements

● Strong TypeScript / Node.js; comfortable in Python where needed.

● Production experience with hosted LLM APIs — prompt design, function/tool calling, structured output, streaming.

● Experience with real-time audio systems: websockets, STT/TTS integration, latency budgets, turn-taking.

● Comfortable with serverless / edge runtimes (Cloudflare Workers, Vercel, Lambda) and their constraints.

● LLM-ops practice: prompt versioning, tracing, evals (Langfuse, Braintrust, Helicone, or similar).

● Solid software engineering fundamentals: testing, observability, incremental delivery.

● Clear written communicator who can work independently across time zones.

Preferred Qualifications

● Prior experience in healthcare, medical education, or other regulated / sensitive-data domains.

● Knowledge of multimodal AI (speech, text, images).

● Contributions to open-source AI projects.

● Experience with HTMX or server-rendered UIs.

● Familiarity with SQLite / D1 / Drizzle ORM or similar typed database tooling.

● Experience operating multilingual speech systems.

● Working knowledge of ML frameworks such as PyTorch, TensorFlow, or Hugging Face Transformers (not required).

● Exposure to fine-tuning, RAG pipelines, or model-agnostic training approaches (not required).

First 90 Days

Within the first three months, we’d expect you to:

● Establish a versioned prompt evaluation harness across our core LLM features.

● Reduce voice pipeline tail-latency by a measurable margin.

● Land a provider migration or model upgrade with a documented quality, latency, and cost comparison.

● Onboard into our LLM-ops tooling and own day-to-day prompt and model changes.

What We Offer

● From £120.00 per day.

● Flexible, fully remote working.

● Opportunity to work on cutting-edge AI applications with real-world impact in education and training.

● A small, focused engineering team with short paths from idea to production.

Job Type: Full-time

Pay: From £120.00 per day

Work Location: Remote