AI Engineer

We are hiring an AI Engineer to build the intelligence layer behind Swirl's agents. This is a hands-on, production-focused role, not a research-only seat. You will design, build, evaluate and operate LLM-powered agentic systems that converse with real buyers, reason over enterprise catalogues, and take real actions like configuring products, structuring finance and booking appointments. You will work directly with the CTO and the core engineering team, own features end-to-end from prototype to production, and have an outsized influence on how the product behaves. If you want to do the most interesting applied-AI work in agentic commerce and see it deployed at enterprise scale, this is the seat.

Responsibilities:

• Build and ship LLM-powered conversational agents that sell across web, chat and emerging AI surfaces, from prototype to production.
• Design and optimise retrieval-augmented generation (RAG) pipelines over enterprise catalogues, inventory, pricing, financing rules and unstructured content (reviews, videos, forums), owning chunking strategy, embedding model selection, hybrid and semantic search, re-ranking, and retrieval evaluation against real buyer queries.
• Implement orchestration for multi-step agentic workflows: tool calling, planning, function execution, and safe action-taking (configuration, finance, booking).
• Engineer prompts and context at scale: prompt design, context construction, and grounding the agent in proprietary context graphs so responses are accurate, on-brand and convert.
• Own model evaluation and quality: build offline and online evaluation harnesses, track quality, hallucination, latency and conversion, and set alerts when behaviour drifts.
• Push model performance where it beats prompting, fine-tuning and distilling smaller models for domain tone, structured output reliability, latency and unit cost, with dataset curation and clear before/after evals.
• Integrate and deploy LLM features into the product backend: robust APIs, streaming responses, caching, fallback and cost control across providers (OpenAI, Anthropic, Gemini and open models).
• Move between research and production: fast prototyping, clean production code, code reviews, and operating what you ship after launch with the team.
• Collaborate with product, design and forward-deployed engineering to translate customer requirements into agent behaviour on tight timelines.

Requirements:

• 1-3 years building software / ML / AI in production, with at least one LLM feature you have shipped and owned end-to-end.
• Strong Python or JavaScript / TypeScript; solid software-engineering fundamentals (testing, version control, clean APIs, debugging in production).
• Hands-on experience integrating LLM APIs (OpenAI, Anthropic, Gemini or open models) into real applications.
• Practical RAG experience with chunking, embeddings, vector databases, re-ranking and retrieval quality tuning, with a way of measuring whether retrieval actually improved.
• Working knowledge of prompt engineering, function/tool calling, and structured outputs.
• Comfort with cloud deployment (AWS preferred), containers, serverless, API services.
• Product sense and ownership: you care about whether the feature actually works for the end user, not just whether the model runs.

Nice-to-have:

• Experience with agent frameworks and orchestration (LangChain, LlamaIndex, LangGraph, CrewAI or custom).
• Exposure to tool-using agents or autonomous commerce workflows in production.
• Fine-tuning, LoRA / PEFT, distillation, or quantisation to cut latency and cost while holding quality.
• LLM evaluation/observability tooling (LangSmith, Langfuse, custom eval harnesses).

Skills

Anthropic, Gemini, JavaScript, LLM, OpenAI, Python, RAG, TypeScript, embeddings, reranking, retrieval, vector databases