Senior Software Engineer, BackendNew

About the role

At Numero , we build software for political campaigns and nonprofits. Since our founding in 2019, we’ve scaled our platform to serve thousands of campaigns ranging from city council seats to US Senate races, and have helped them collectively raise over $10B in contributions. Our platform is an all-in-one system comprised of a donor CRM, broadcast email + texting tools, and payment processing pages that our clients rely on for their work every day. As our product is the platform that our users operate out of every day, this also means that when we build and launch new products, we do so to all of our users, at scale 🚀. We’ve experienced tremendous growth over the past several years and are hiring experienced engineers to help us scale our backend infrastructure and to support the development of several new products on our roadmap. In joining our team, you will help shape our team’s culture and have a substantial impact on the success of our company 📈 – come join us! Who we’re looking for The work at Numero is challenging because of the nature of what we build: systems where a subtle edge case or an inefficient query is felt by every client, on every campaign, during the highest-traffic moments of an election cycle. We're looking for someone who finds that kind of nuance and scale genuinely fun to untangle – and who has the experience to do it well. Deep experience building and scaling user-facing backend systems. Ideally 6+ years, with a meaningful portion spent responsible for production systems with significant traffic. Leveling and salary will be commensurate with experience. A deep, instinctive understanding of how code behaves at scale. You think naturally about the code paths that get hit by millions of requests, and you reason about cost before it shows up in a flame graph. Real fluency with SQL and relational data at scale. You can anticipate bottlenecks and use optimizations like indexing and batching to write efficient, scalable queries. You can analyze performance metrics and monitoring after shipping to ensure things are working. A bias toward building things the right way. You architect patterns that scale well even when day-one usage looks small – because you've seen how "we'll fix it later" rarely survives contact with growth. Hands-on experience building with AI – and a hunger to get sharper at it. You already use AI tools in your day-to-day engineering and want to deepen that craft as agentic coding matures; we're investing heavily here and want teammates pushing the frontier with us. Excellent written and verbal communication and high EQ. On a team this small, how you reason out loud and collaborate matters as much as what you ship. You are located in the continental USA (ET, CT, MT, PT) and are willing to travel 1-2x a year for company retreats and/or team offsites (usually held in California). We're a tight-knit team, energized by the challenges in our work because of how far its impact reaches – what we build touches every campaign and nonprofit on the platform within minutes of deploying. That makes the problems we tackle genuinely meaningful: the things you ship move the needle for clients running some of the most consequential campaigns in the country. What we’re working on Every feature of Numero is built with best-in-class engineering and design, in close coordination with our clients. Much of the work carries real complexity – money movement, compliance, and querying at scale – and because we're an all-in-one platform, most of it reaches every client the moment it ships, often at a scale of millions of requests and billions of records. A sampling of what's on our plate right now: Numero Agents – a suite of AI agents that let clients drive the platform in plain language instead of doing the work by hand, each equipped with the context, tools, and domain knowledge that make it sharp at its job. There's deep, ongoing work here across the shared intelligence and orchestration layer that powers these agents, the retrieval that grounds them, and the tooling they use to act – all built on our own Ruby LLM tooling. Scaling Numero QL – Numero QL serves millions of requests a day, filtering billions of records across multiple databases in realtime. Moving our highest-frequency queries onto OpenSearch is one technique we're using to push performance further, and we're actively exploring others – including traditional approaches like database partitioning and sharding. Tandem Fundraising – one of our most-requested features ever: splitting a single contribution across multiple recipients, with precise handling of fees, compliance reporting, refunds, and a long tail of edge cases – across both Stripe and PayPal. Payment Processing Optimization – This is our fastest-growing vertical, and we're focused on getting more supporters across the finish line. Two big levers: remarketing – automatically following up on abandoned checkouts by email – and letting returning donors securely save their information and reuse a payment method for faster, lower-friction giving. At our scale, every percentage point of conversion compounds across millions of donors. Communications – Building deeper integrations to support calling, texting, and the real-time infrastructure behind them – transcription, syncing communications data into the CRM, and conversational AI that handles the follow-up busywork for clients. Custom Fields – letting clients define their own fields to store and query any type of data in their CRM, with configurable data types, format validation, and performant querying through Numero QL. Our roadmap runs on a quarterly horizon with a lot of built-in flexibility: we follow our clients' needs and our fastest-growing revenue segments, always optimizing for what's best for them and for the company. And because we're a small team, we rotate across these product lines, so you'll get exposure to a wide range of the work. How we work with AI AI is core to how we build at Numero – not a bolt-on. It runs through every coding project we take on, and we've invested heavily in the tooling and workflows around it. We expect everyone here to be a sophisticated operator of these tools, and to help make them better. In practice, that looks like: Planning and estimation. We map out projects using Plan Mode with Claude and Codex before writing code – surfacing edge cases, data-modeling decisions, and architectural tradeoffs early. We'll sometimes have an agent one-shot a feature and open a throwaway PR; we don't ship that code, but it's a fast way to expose the real dependencies and gotchas that shape how long the work will actually take. Pair programming with AI. We work alongside tools like Claude and Codex throughout development – to move quickly and make fewer mistakes, propagate a pattern, write tests, and answer questions or debug as we go. Keeping their output consistent with how we build is where our Skill files come in. Skill files as institutional memory. Our conventions and hard-won lessons live in Skill files in version control – reviewed through PRs like any other code, and scoped by domain ( Numero QL, background jobs, payments, and more). When an agent produces something off-pattern, we don't just fix it; we encode the lesson back into a Skill file so it never recurs. Review and debugging. Every PR runs through automated review with Greptile and Cursor Bugbot, with deeper passes from Claude and Codex – so authors catch and fix bugs before requesting review, and human reviewers can get up to speed on a change without combing through every line and file before they approve and merge. For bugs, where the real work is finding the root cause, we lean on Sentry Seer and repo-aware Claude/Codex to investigate, which has sharply cut our debugging time. Maintenance and migrations. Package upgrades and large, codebase-wide migrations are some of our highest-leverage AI work – agents often hold more context on the migration path than we do. Agentic projects. Alongside their core work, every engineer runs an agentic project : completing a real feature – e.g. a new export format, a third-party integration – entirely end-to-end using agents (planning, building, testing, and breaking the work into small, incremental PRs that match how we ship production code). These are real, shippable features we want built but deliberately hold to a looser timeline – so the work happens in normal hours, with room to do it agent-first – the point being to sharpen how we work with agents and push their output toward our preferred patterns, codifying a new skill-file rule whenever they miss one. We also debrief our AI wins and misses in every sprint retro, so the whole team compounds what each of us learns. For day-to-day AI coding, we each work with a mix of tools – Claude Code, Cursor, and Conductor among them – and engineers are free to reach for whatever fits them best. If you've been pushing the edge of how AI fits into real engineering work – with opinions on where it shines, where it doesn't, and where it's headed – you'll fit right in. What we value Here are some of the concepts that we value as a team and hope resonate with you: Consistency and clarity. You've seen how much a consistent, legible codebase pays off on a small team – in shared understanding, developer happiness, and velocity. It's also what makes our AI tooling work: clear patterns that we can codify in skill files so that our agents produce code we're likely to approve. A shipping-first instinct. You're driven to get real value into customers' hands quickly – often by getting a first or second draft out fast with agents, then refining it into a production-ready PR that doesn't sacrifice quality or resilience. Follow-through and good hygiene. You finish what you start, and you leave the codebase cleaner than you found it – paying down tech debt as you go, often pointing agents at existing code to standardize on better patterns. Eagerness to learn – and to shar