About Our Client:
Our Client builds institutional memory for investment firms and family offices. The judgment, decisions, and conversations that make a firm what it is mostly live in people's heads — and disappear when those people move on. We change that, working on top of the tools each firm already uses to turn how they operate into something the team can use, allocators can see, and the next generation can inherit.
The Role
This is a frontline operations role for an AI-native consulting practice. The founder owns each client engagement and sets the architecture. You keep the systems running and the clients well-served.
As AI lets very small teams deliver work that used to require very large ones, someone has to make sure the prompts, knowledge bases, capture routines, and recurring deliverables behind that actually work as intended.
What you'll do:
- Frontline support: Be the first point of contact for client teams. Take in requests, respond, escalate to the founder when needed.
- Diagnose the problem: Most requests won't come with clear specs. Figure out what the client actually needs and what the response should look like.
- Update the tooling: Once you know the right response, go into Notion, Linear, Slack, Google Workspace, Claude, Fireflies, etc. and make it real. The work is configuration and curation.
- Turn messy material into useful artifacts: Transcripts, email threads, scattered documents — read them, find the structure, produce a summary, decision log, brief, or procedure.
- Steward the knowledge base: Keep each client's institutional memory layer organized, current, and trustworthy. Catch staleness, gaps, and drift before they compound.
- Monitor quality: Review AI-assisted outputs on a regular cadence. Fix wrong-voiced or thin output and feed corrections back into the templates.
- Document the system: Write the procedures and reference materials that let a client's team operate on their own.
- Onboard new clients: Stand up tooling, ingest context, build initial structures, get routines running. Each onboarding makes the next one faster.
AI operations didn't exist as a job a couple of years ago so we care more about how you're wired than what's on your résumé.
- You're wired for correctness: You also can't comfortably walk away from something that's half-finished or quietly broken. You notice when an output is subtly wrong, thin, or off-voice.
- You learn by doing: You don't wait for perfect documentation; you take a stab, and when you get stuck you come back with a precise account of what you don't understand.
- You teach yourself the frontier: You already poke at new tools, features, and techniques on your own time to see how far you can push them because learning is fun for you.
- You want to own outcomes: You're the person who can see the fix and is frustrated by not having the authority to deliver. You want the accountability — not the title, the responsibility — and you understand that owning something means defining what success looks like and being measured against it.
- You can understand what people actually need: Requests arrive messy and underspecified. You're good at translating "the client said X" into "here's what actually needs to happen," and triaging across tools and teams to make it real.
- You do not need a finance background or a degree: You do need to be smart, curious, easy to work with, and genuinely interested in the problem.
A Litmus Test (Read Before you apply):
This role is about operating and wielding advanced AI systems — not building them. You will not be asked to code or to derive any math. But you do need enough conceptual fluency to know what these systems are doing and when they're going wrong.
Before you apply, spend an afternoon — call it four hours — getting the gist of three things (YouTube is fine):
Before you apply, spend an afternoon — call it four hours — getting the gist of three things (YouTube is fine):
- Evals and traces — how we can measure whether an AI system is actually doing its job, and how we follow what it did, step by step.
- GraphRAG / Hybrid RAG — how systems retrieve the right memory and context.
- Auto-research loops — e.g. Andrej Karpathy's framing, or "Codex goals" and "Claude dreaming" as different takes on the same core idea.
- You don't need to understand how it works. The bar is simpler and harder: can you grasp the value, and picture how you'd operate and wield these in practice?
If after an afternoon you can honestly say "I don't know how it works under the hood, but I get what it's for and how I'd use it" — you'll thrive here. If it's all still fog after a real attempt, this probably isn't the right seat, and that's genuinely okay.
Originally posted on