About Calvis
Calvis is an AI-native physical security platform. Traditional security never learns — guards miss events, patrol routes stay static, and customers find out whether a shift went right only after it's over. We fix that by uniting many licensed security agencies into one intelligent network, then wrapping every shift in an AI Copilot and a 24/7 human Overwatch team. The result: a security operation that gets smarter with every shift — trusted by businesses across warehousing, distribution, manufacturing, retail, and events.
About the role
A growing share of Calvis runs on AI agents: copilots that supervise live shifts and talk to guards, agents that support sales and operations, and automations that keep hundreds of accounts healthy. You'll design and build these agents: the prompts, the tools they call, the memory and guardrails around them, and the evaluation that proves they work.
What you'll do
- •Build and ship LLM-powered agents that operate in production, around the clock
- •Design tool interfaces, memory, and orchestration for multi-step agent workflows
- •Decide what agents may do on their own and when they hand off to a human, then enforce it
- •Build the evals and monitoring that tell us how well agents actually perform
- •Drive down cost and latency while quality goes up
What we're looking for
- •Strong software engineering fundamentals (Python preferred). This is an engineering role, not a prompt-writing role
- •You've built something real with LLMs: agents, tool use, RAG, or eval pipelines
- •Unattached to any particular model or framework; you pick whatever ships
- •Comfortable owning systems that interact with real customers and employees