The posting
About Conduit
Conduit is an AI agent platform purpose-built for hospitality. Hotels, vacation rentals, and resorts deploy our agents across guest-facing communication and internal operations, running millions of workflows a day.
Hospitality is 10% of global GDP and it turns on customer experience. Resolving a guest request immediately is most of that experience, and resolving one means moving across several tools and between several people. That coordination is the work, and it is what our agents do.
Role: Member of Technical Staff
There is no hierarchy on the engineering team and everyone can work anywhere in the stack. People end up specializing in the product lines they gravitate toward, but that happens on its own rather than by assignment. We want the engineering org as flat as we can keep it.
Why this role is different
- Free rein with AI agents, and full ownership of what comes out. Nobody tells you which tools to reach for or how much of the work to hand off. Where to use an agent and where to do it yourself is your call, and how well you make that call is what we evaluate you on.
- You will touch most of the stack and end up owning parts of it. You work across the codebase instead of in one lane. The modules you go deep on become yours, and the team depends on you for them.
- The design decisions are yours to make. You choose the approaches, and the ones that hold up stay in the platform for decades. That is real weight to carry and it is most of why the role is interesting.
- You are directing a fleet of agents. You sketch the work, design it carefully, and the agents execute against that design in parallel. You get a lot of output for each unit of work you put in, and getting good at running that is a skill that follows you everywhere else.
What the work looks like
- Response generation pipeline. The router, orchestrator, and synthesizer that classify intent, assemble context, select tools, generate a response, and decide whether to send it or escalate.
- Knowledge retrieval. Models ground their answers in whatever context they are given, so reliability comes down to whether that context is correct, complete, and current. Customer data is usually none of those. You build the retrieval, ranking, and validation that makes the output trustworthy anyway: a hierarchical knowledge base with scope-based queries, contradictory docs, stale articles, and the edge cases that live in one rep’s head.
- Tool orchestration. The agents modify reservations, dispatch maintenance, send confirmations, and close the loop. You design the tool-use and function-calling layer that runs multi-step workflows end to end, including what happens when a third-party API goes down mid-conversation.
- Evaluation and testing. The eval framework that measures quality against real conversations and catches regressions before customers see them.
- Observability. The trace that shows what the agent retrieved, what it considered, what it ignored, where confidence dropped, and why it chose what it did. That trace is what lets a CX team trust the system and improve it.
- Controls for people who are not engineers. The people running these agents day to day are support leads. You make the agent’s behavior legible, testable, and adjustable by someone who has never written code.
- Voice. Real-time conversations with transcription and the same tool use as text, under tighter latency.
- Speed and cost. Model routing, caching, prompt optimization.
- Infrastructure, data pipelines, integrations, and API work. The systems underneath all of the above.
Who you are
- You have shipped production systems that real users depend on.
- You are comfortable anywhere in the stack and pick up whatever the problem needs.
- You think about retrieval as seriously as generation.
- You are pragmatic about models. Sometimes the right answer is a frontier model, sometimes a small fine-tuned one, sometimes a regex.
- You care about evaluation. You do not ship a prompt change without knowing what it does across a thousand real conversations.
- You want a small team where the thing you push changes the product that day.
Compensation and Benefits
- $130-220K base
- 0.20%-1.20% equity
- Comprehensive health insurance and 401k match
Interview Process
- First call (20 min) with the founders. What you have built and how you think about running AI systems in production. No prep needed.
- Technical interview (45 min). An AI systems problem: debugging a retrieval pipeline returning wrong context, designing tool use for a multi-step workflow, or working out why a response went sideways on a real conversation. No leetcode.
- Optional agent coding session (5 min video). Record yourself working through a problem with an AI coding agent. Not required, and it tells us a lot about how you work.
- On-site work day in San Francisco. Spend a day with the team, scope a small feature together, build it and ship it. Travel and time are compensated. This role is onsite in San Francisco (not remote). Candidates must be authorized to work in the US; we do not sponsor visas.
- References and offer.
