Responsibilities

This is not a chatbot or a demo. The agent acts on real money, on behalf of real executives, with real venues on the other end of the thread. It runs in production today for reservations and private dining. Our strategy depends on carrying more of the work with every request: more kinds of entertainment, more cities, more of the workflow, without growing headcount to match. You will set how we get there.
Own the agent system. Ande runs on a root concierge agent with specialist subagents for sourcing, comparison, and venue coordination, on a self-hosted agent harness behind queued turn workers. You’ll own its architecture, model choices, and upgrade path, and make sure it holds up under latency budgets, provider failover, and real-world load.
Build tools the agent can trust. The agent has about 75 typed capabilities that let it search, hold, book, pay, and message. Each one runs as the acting user and respects who they’re acting for, such as an assistant booking on behalf of an executive. Consequential actions become proposals that the booker confirms. You’ll decide what the agent is allowed to do, and how to make it safe.
Make evals the gate for every change. You’ll build scenario suites, graders, and LLM judges calibrated against human labels. They replay from recorded runs on every PR and run in full nightly. Your goal is evals trustworthy enough that a model upgrade or prompt change can ship on the strength of the number alone, and cheap enough to run all the time.
Power venue correspondence. Our agent negotiates with venues over email. It reads replies, pulls out offers, minimums, deadlines, and payment requests, and knows when to push, when to wait, and when to hand off to a person. You’ll make sure all of it is resistant to anything a third party writes into a thread.
Build memory and the supply learning loop. You’ll shape what the agent remembers about a company, its people, and its venues, with clear rules on what’s shared and what stays private. You’ll also build the loop that turns every venue conversation into structured supply data, like which room actually fits 14 or what a venue will do on a Tuesday, so each recommendation is better than the last.
Ship fast, reliable judgments. You’ll build typed classifiers for consent, triage, ranking, and urgency. Each has labeled thresholds, a timeout, and a fallback, and each rolls out from shadow mode to live.
Set the bar for building with AI. You’ll define how Ande builds with AI: agent behavior as versioned config, product teammates proposing changes through evals, and AI coding agents as the default way the team ships. You’ll mentor engineers and partner closely with our AI product manager.

Qualification

Education and experienceProduction evalsSystems engineeringSound judgmentAI-native workflowClear communicationYou've built agent harnessesExperience with searchExperience with marketplacesWhy This Role MattersThe Ande Agent is how we scaleThe stakes are real

Required

Education and experience: a degree in Computer Science or Engineering, or equivalent practical experience. You have 5+ years building production software, including 2+ years shipping LLM agents to real customers (tool use, multi-step planning, structured outputs, long-running conversations).
Production evals: you’ve owned evals in production, from writing them to finding failure modes in real traffic and improving quality turn by turn.
Systems engineering: strong TypeScript and backend skills, including queues, concurrency, idempotency, and how distributed systems fail.
Sound judgment: you know where a model belongs and where deterministic code does, and you have the discipline to make agent actions safe, auditable, and reversible.
AI-native workflow: you use AI coding agents every day. Our codebase is built to be worked on by them, and we expect you to raise that bar.
Clear communication: you can explain agent behavior and trade-offs to product, operations, and customers.

Preferred

You've been a founding or early engineer and know how to balance craft, ownership and speed.
You've built agent harnesses, orchestration layers or tool brokers, not just prompts.
Experience with search, ranking or recommendation systems.
Experience with marketplaces, booking or payments, where a wrong action costs money.
Experience defending agents against prompt injection from untrusted content.
Why This Role Matters
The Ande Agent is how we scale. Every request it handles end to end is work we don’t have to hire for, so how far it can take us depends on the person who owns its intelligence.
The stakes are real. The agent moves real money for executives and negotiates directly with venues, so a mistake costs customers money and costs us venue relationships. Every venue conversation also makes the agent smarter. That learning loop is an advantage competitors can’t easily copy.

Software Engineer, AI (Early Career) at Ande · LumiJobs

Software Engineer, AI (Early Career)

Responsibilities

Qualification

Required

Preferred

Company

Ande

More roles at Ande