Software Engineer, AI

Ande•New York City, NY

23h•Onsite

About The Position

Ande is hiring a Software Engineer, AI to own the intelligence behind its production Ande Agent, a concierge system with specialist subagents that turns customer requests into confirmed, paid, on-policy entertainment bookings. The agent currently handles reservations and private dining and works with real customers, executives, and venues. The role will shape how the agent expands to support more entertainment types, cities, and workflow steps without a proportional increase in headcount. The engineer will own the agent system’s architecture, model choices, and upgrade path, and ensure it operates reliably within latency budgets, through provider failover, and under real-world load. They will build safe, typed tools for agent actions, including search, booking, payment, and messaging; create production evaluation systems to test changes; and improve the agent’s handling of venue correspondence, memory, supply data, and operational judgments. The role also includes defining how Ande builds with AI, mentoring engineers, and partnering closely with the AI product manager. This is an in-office position in New York City, four days per week. Ande is an equal opportunity employer and welcomes applications from all qualified candidates.

Requirements

A degree in Computer Science or Engineering, or equivalent practical experience.
At least 5 years of experience building production software.
At least 2 years shipping LLM agents to real customers, including tool use, multi-step planning, structured outputs, and long-running conversations.
Experience owning production evaluations, including writing evaluations, finding failure modes in real traffic, and improving quality over time.
Strong TypeScript and backend engineering skills, including queues, concurrency, idempotency, and distributed-systems failure modes.
Sound judgment about when to use models versus deterministic code, and discipline in making agent actions safe, auditable, and reversible.
Uses AI coding agents every day and can help raise the bar for an AI-native development workflow.
Clear communication skills to explain agent behavior and trade-offs to product, operations, and customers.

Nice To Haves

Experience as a founding or early engineer, balancing craft, ownership, and speed.
Experience building agent harnesses, orchestration layers, or tool brokers.
Experience with search, ranking, or recommendation systems.
Experience with marketplaces, booking, or payments, where incorrect actions have financial consequences.
Experience defending agents against prompt injection from untrusted content.

Responsibilities

Own the architecture, model choices, and upgrade path for the root concierge agent, specialist subagents, self-hosted agent harness, and queued turn workers.
Ensure the agent performs reliably within latency budgets, with provider failover and under real-world load.
Build and govern typed agent capabilities for searching, holding, booking, paying, and messaging, with appropriate user permissions and confirmation proposals for consequential actions.
Create scenario suites, graders, and LLM judges calibrated against human labels; run evaluations on every pull request and in full nightly, and use results to guide model and prompt changes.
Make venue correspondence handling robust, including extracting offers, minimums, deadlines, and payment requests, deciding when to negotiate, wait, or hand off, and resisting prompt injection from third-party content.
Shape agent memory about companies, people, and venues, with clear rules for shared and private information.
Build a learning loop that converts venue conversations into structured supply data to improve future recommendations.
Build typed classifiers for consent, triage, ranking, and urgency, with labeled thresholds, timeouts, fallbacks, and staged rollouts from shadow mode to live.
Define Ande’s approach to building with AI, including versioned agent behavior configuration, eval-based product changes, and AI coding agents as a default workflow.
Mentor engineers and partner closely with the AI product manager.