Senior AI Engineer

About us

============

Ruby Labs is a leading tech company that creates and operates innovative consumer products. We offer a diverse range of opportunities across the health, education, and entertainment industries. Our innovative teams are driving the future of consumer-led products, and we're always looking for passionate individuals to join us. Learn more about our story at: https://rubylabs.com/about-us/

About the role

==================

We are looking for a Senior AI Engineer to own and scale the AI layer of one of our core D2C subscription products. You will be fully responsible for production LLM features and the browser automation built around them end to end — prompt systems, structured outputs, evaluation, observability, and model selection — from first experiment to live release, with quality, reliability, and cost as your accountability.

This is a high-impact role suited for someone who decides with data, not intuition. Unmeasured prompt changes, untraced failures, and model swaps without benchmarks are not acceptable. You work in a modern TypeScript stack — Next.js, Cloudflare Workers and Containers with Hono, Hatchet task queues, Drizzle on Supabase Postgres, Vercel Blob. Browser automation on Browserbase and Playwright is roughly half of the job.

Key Responsibilities

========================

Prompt Systems & Structured Outputs

• Design dynamic prompt templates with conditional logic and context injection that maximise generation quality
• Enforce predictable outputs with JSON mode, function calling, Zod/JSON schemas, validators, and repair loops
• Own complex LLM workflows (LangChain, LlamaIndex, or custom) that integrate cleanly into application logic

Browser Automation

• Build and maintain browser automation on Browserbase Functions and Playwright that traverses web flows reliably
• Classify screens with a model inside each traversal, and decide what the model handles versus deterministic code
• Ship automations as frozen, published workers — versioned, reproducible, and safe to roll back
• Orchestrate runs through Hatchet with retries, timeouts, and failure traces that explain every broken run

Evaluation & Experimentation

• Build evaluation pipelines — test datasets, LLM-as-a-judge, RAGAS-style metrics — and score responses and screen classifications in Langfuse
• Trace complex LLM chains to find bottlenecks in cost, latency, and context-window usage
• Run A/B experiments across models and prompts through an AI gateway (OpenRouter, Cloudflare AI Gateway); release only on quantitative benchmarks
• Re-evaluate models as new architectures emerge; fine-tune when domain requirements demand it

Production Ownership

• Take AI features from experiment to live production on Cloudflare Workers and Containers, and stay accountable after release
• Manage rate limits, fallbacks, and model cost so features stay reliable at scale
• Resolve production incidents in owned AI and automation…