About Rivet
Rivet is an AI-native accounting firm. In less than 3 years, we've built a T500 firm in the US, combining corporate and personal income tax, sales tax, R&D credits, and bookkeeping into a single, all-inclusive experience that eliminates the worst parts of tax season for real businesses and the people who run them.
Our platform is designed to feel invisible: it guides clients through their tax to-dos, automates the drudgery for preparers, and gets out of the way when it's not needed. We measure success in time saved and returns completed, not vanity usage metrics. We run Rivet (the firm) on Rivet (the platform) because owning the stack makes it possible to deeply embed AI into every step of our workflows, far more than our competition will ever be able to relying on third-party/off-the-shelf tooling.
Our team of 50 supports more than 1,000 of the most innovative companies in the US — including Cursor, Cognition, and ModRetro — and their executives, investors, and employees.
And while we may have just gotten started a little over two years ago, this isn't our first rodeo:
Nick Abouzeid (cofounder & CEO) comes from Ramp, where he ran three of their Growth teams from $100M → $600M in ARR
Kyle Christenson (cofounder & Head of Tax) comes from the IRS, where he spent six years as an auditor before leaving to start his own firm over a decade ago
We're well-funded, growing quickly, and sustainable. We've raised over $7M to date, from investors like XYZ Ventures, Susa Ventures, Haystack Ventures, and 40+ executives from Ramp, Anduril, Physical Intelligence, & more.
Why this role is different
Tax software fails quietly. A number lands in the wrong box, nobody notices until a notice arrives, and by then it's a client's problem. The backend is where we decide that doesn't happen.
Ownership from first commit to production. Engineers design, build, ship, and maintain their work — no handoffs, no middle-management mazes.
Decisions happen at the edge. Engineers talk to clients and tax preparers directly. We optimize for speed and context, not committee consensus.
Clear is kind. We surface problems early, talk about tradeoffs openly, and don't sugarcoat. We're kind by being honest, not by being vague.
Pragmatic stacks. We pick tools that solve real problems. If it doesn't help us ship faster, safer, or better, we don't use it.
Skin in the game. We ship things we'd be proud to put our name on — and if something breaks, we own it and fix it fast.
Our stack
Six engineers today, five languages. Our platform and infrastructure are Go end to end:
Platform: Go, as a modular monolith — Connect RPC over protobuf, Go Fx for wiring, sqlc and goose against Postgres, River Pro for background jobs, Watermill for pub-sub over RabbitMQ.
Frontend: React and TanStack, deployed to Cloudflare Workers.
AI and eval: Python, in their own repos — ModelBridge (a gRPC service for AI building blocks) and TaxBench, our internal eval suite.
Legacy integration: a C# Windows agent that drives Drake, one of the tax platforms our industry never built APIs for.
Infrastructure: AWS — ECS Fargate, Amazon MQ, AMP, Loki, Tempo — provisioned by Pulumi, itself written in Go.
Around the edges: Buildkite for CI, OpenTelemetry throughout, Doppler for secrets, Linear for tracking.
What you'll do
Own modules in the Go monolith end to end — schema and migrations through Connect RPC surface, including the background jobs and pub-sub flows behind them.
Automate systems nobody meant to be automated. Our industry never built APIs. We drive a desktop tax platform through a C# agent, parse filings that arrive as scanned paper, and reconcile ledgers that disagree with each other. This is most of the interesting work.
Put AI on the critical path. Wire ModelBridge into real workflows and hold it to TaxBench, our eval suite — we treat eval as an asset, not a side project. A model that's right 95% of the time is a design problem, not a launch.
Own what you ship in production. Tracing, alerting, on-call, and the load curve of a firm whose traffic triples in March. April 15th doesn't move.
Work directly with preparers. The people who'll catch your bug are on Slack, and they'll tell you exactly what happened.
We're fully distributed — but if you want an office, we'll help you find one through our investor network.
What makes you a fit
4+ years building backend systems, including 2+ years writing Go in production
Real Postgres depth — schema design, migrations against live data, and knowing why a query got slow
Have built and operated asynchronous systems: background jobs, queues, retries, and the failure modes that come with them
Bias toward correctness in a domain where a wrong number has consequences, and the judgment to know when it doesn't
Comfortable operating in a fast-changing, early-stage environment
Excellent verbal and written communication skills
Nice to haves
Connect RPC, gRPC, or protobuf-first API design
Event-driven architectures on RabbitMQ, Kafka, or similar
Infrastructure as code — Pulumi, Terraform, or CDK — and comfort owning your own deploys
Shipped LLM systems into production paths and evaluated them properly, not as prototypes
Built tax-adjacent software — fund administration, payroll, accounting platforms, filing systems
Enjoy getting creative with APIs and systems that were never meant to be automated
What we offer
Benefits and meaningful equity
A fully remote team, plus twice-yearly company offsites
A well-funded, fast-moving environment with minimal layers of management
A world-class client base that values your time and pays for the premium experience
Problems that are unusual — nobody else is writing daemons against 1990s tax software to feed a modern agent stack
We encourage you to apply even if you don't meet every qualification. Strong candidates rarely check every box.