We're an applied AI engineering team. Other teams bring us hard problems; we figure out how to solve them with AI — right now, mostly with agents. We do the full loop: research the space, design an approach, prototype fast, and validate whether it actually works. We move in weeks, not months.

Depending on the project and its stage, you might:

• Scout the landscape — existing tools, papers, community threads, open-source approaches — and build a clear map of what's been tried and what's working.

• Design agent architectures for problems that don't have a known solution — deciding what tools, memory, and coordination patterns might work before anyone has proven they do.

• Build prototypes and run targeted experiments — enough to know if an approach works, not a full benchmark suite.

You'll work on real problems with teammates who've been deep in this space.

• You build AI agents on your own time — and you keep at it. You follow the space — new models, frameworks, papers, community threads — and when something drops, you try it.

• You've hit the walls that don't show up in tutorials — context that fills with noise, memory that drifts, tool calls that break in ways no framework warns you about. You've seen multi-agent setups that look clean on a whiteboard and fall apart at runtime.

• You write code you can explain and debug — not just prompt until it works. You use AI tools to move faster, but you know when to trust them and when not to.

• You understand how LLMs work below the API layer — attention, tokenization, how sampling shapes output, why a model that "knows" something still fails to use it. When something breaks, you reason from mechanics, not just tweak and retry. You can design an eval that measures what you intend — and catch when it doesn't.

If most of this sounds familiar from your own projects — not just from reading about them — we'd like to hear from you.