AI engineer
We’re looking for someone to turn raw language models into intelligent, reliable product features.
That means prompts, tools, retrieval strategies, evaluations, and everything else that makes LLMs behave. You’ll work directly with the founders to design how AI shows up in the product—from quick prototypes to robust production flows. This isn’t a research role. It’s about making applied AI real, useful, and fast.
This is the only AI-focused hire for now—and a defining one. You’ll shape how models are used, how they’re evaluated, and how they’re integrated into the core of the product.
What you’ll do
This role is about making intelligence usable—designing the reasoning layer that sits between models, tools, and the user.
Own the prompt architecture: few-shot, zero-shot, modular, composable—whatever it takes to get reliable output
Design and build RAG pipelines that ground model outputs in real data
Orchestrate how the model interacts with tools, APIs, and other systems—function calling, fallback logic, and flow control
Set up evaluation frameworks to measure quality, hallucination risk, and overall reliability
Work with product and engineering to prototype fast, test with users, and turn insights into robust building blocks
Handle messy data, ambiguous use cases, and weird model behavior—debugging and refining as you go
Help shape our long-term AI strategy: model selection, fine-tuning, vertical adaptation, and more
Treat prompts, tools, and evals as code—not just experiments
Who you are
You’ve worked directly with LLMs. You think in systems, not just prompts. You’re excited by the challenge of turning unpredictable models into dependable features.
You’ve used OpenAI, Claude, or open-source models in real product contexts
You understand token behavior, context limits, and how to manage cost vs. performance
You’ve built chains, flows, or agents—and made them stable enough to ship
You’ve worked with retrieval systems, embedding models, or custom function routing
You’re comfortable writing Python, moving fast, and debugging weird outputs at the edge of what’s possible
You think deeply about model evaluation, not just capability—what good looks like, and how to measure it
You’re product-minded: you care about how the AI feels and how it fits into the user’s workflow
You don’t see prompts as glue. You see them as architecture.
You’ve got 5+ years of experience working at the intersection of ML, product, and systems—and you’ve shipped AI features that actually worked in the real world.
You’ll be the person who makes the AI trustworthy. And you’ll shape what this becomes—just as much as any line of code or pixel on screen.
How to apply
Show us what you’re proud of — projects, code, writing, anything that reflects how you think and build. Format’s up to you. Send it all to jobs@iconic.works
