Founding Workflow Engineer

Founding Workflow Engineer, Agent Operations Location: New York, NY (hybrid, three days in office) or remote within US time zones Employment type: Full-time Compensation: $165,000 – $205,000 base, plus equity Team: Agent Runtime & Client Workflows About Nummo Nummo builds AI agents for recruiting and staffing firms. Our desktop app and agent runtime handle the repetitive work that fills a recruiter's day: reading inbound email, updating candidates in the applicant tracking system, scheduling interviews, writing meeting notes into the right record, and drafting follow-ups for review before they go out. Our clients run executive search, contract staffing, and in-house talent teams. Many of them work mostly from WhatsApp and their inbox, so the product has to fit how they already work. We are a small team. Every engineer ships to production, talks to clients, and owns outcomes rather than tickets. We care about simple systems, honest verification, and software that recruiters can trust with their candidate relationships. The role We are hiring a Founding Workflow Engineer to own the layer between our shared agent runtime and the specific ways each client works. You will take a client request, such as "move every candidate who accepts an offer into the onboarding pipeline and notify the hiring manager," and turn it into a general tool in shared code plus a client-specific prompt and configuration. You will decide what belongs in the platform and what belongs to one firm, and you will prove that the result works with scenario evals and real runs in the app. This is a hands-on engineering role with direct client exposure. You will sit in on calls with recruiting leads, read through their real workflows, and come back with a design that serves them and the next ten clients too. What you will do Design and build agent tools that integrate with applicant tracking systems, calendars, email, note-taking apps, and messaging platforms, starting with Workable, Google Calendar, Gmail, Notion, and WhatsApp. Write and maintain client workflow prompts, and keep client names, columns, and step rules out of shared code. Write scenario evals that describe a workflow in plain English, run them against base and head builds, and use the results to decide whether a prompt change ships. Debug production agent runs using traces, Temporal workflow history, and read-only database access, then fix the root cause instead of adding a server-side guard. Build workflows that finish every action the available facts support, hold the rest with a clear explanation of what is missing, and resume when the missing information arrives. Work with our desktop engineers on how automations, tasks, and approvals appear in the Electron app, so recruiters can see what the agent did and why. Help non-technical teammates write their own evals, and review their work so the eval suite stays trustworthy. Join a lightweight on-call rotation for client-facing automations. What we are looking for Required 4+ years of professional software engineering experience, with significant time in TypeScript and Node.js. Experience building production systems on top of LLMs: tool calling, prompt design, structured outputs, and evaluation. Comfort with Postgres and with reading and writing SQL directly. Experience integrating third-party APIs with real-world complications such as OAuth, rate limits, webhooks, pagination, and inconsistent data. A habit of verifying your work in the running product, not only in unit tests. Clear written communication. Much of our design work happens in PR descriptions and short written proposals. The judgment to say "this is a prompt problem, not a code problem," and the reverse. Nice to have Experience with Temporal or another durable workflow engine. Background in recruiting, staffing, or HR technology, or close work with recruiters. Experience with Electron or desktop application development. Familiarity with eval tooling such as Braintrust, or with building your own eval harness. Experience at an early-stage startup where you owned a product area end to end. How we work One runtime, many clients. Automations, chats, channels, and tasks all run through the same agent loop. Client differences live in prompts and configuration, not in branches inside the runtime. Delete first. We would rather remove a feature than maintain one nobody uses. A smaller diff is usually a better one. Prove it. A change is done when the real flow works in the app and the evidence names the exact commit. A green test suite is not proof. Humans own the high-risk calls. Agents and engineers move quickly on most changes. Authentication, external sends, billing, and releases get explicit human sign-off. No polling, no guesswork. Our desktop app updates in real time from database change events, and our agents never invent facts to finish a task. What success looks like In your first 30 days, you will ship a fix to a live client workflow, write your first scenario eval, and run a full automation end to end against a test account. By 90 days, you will own the workflow library for at least two clients, have shipped one new general-purpose tool that more than one client uses, and have cut down a recurring class of agent mistakes with a prompt change backed by evals. By the end of your first year, you will have shaped how Nummo onboards new clients, so that a new firm's workflows can go live in days instead of weeks. Benefits Competitive salary and meaningful early-stage equity Medical, dental, and vision insurance with 100% of employee premiums covered 401(k) with a 4% match Flexible paid time off with a 15-day minimum we actually enforce $2,000 annual learning and conference budget Home office stipend and the latest MacBook Pro Quarterly in-person team weeks Interview process Intro call (30 minutes) with a founder about your background and what you want next. Technical conversation (60 minutes) about a system you have built, the tradeoffs you made, and what you would change. Take-home or paired exercise (about 3 hours, paid): turn a sample client request into a tool, a prompt, and an eval. Onsite or virtual final (half day) with the team, including a client-workflow design session and a values conversation. References and offer. We usually finish the process within two weeks. Equal opportunity Nummo is an equal opportunity employer. We do not discriminate on the basis of race, religion, color, national origin, sex, sexual orientation, gender identity, age, disability, veteran status, or any other legally protected status. If you need accommodations during the interview process, tell us and we will arrange them.

Requirements

na