Shovels is the prediction engine for the built world. Everything built starts with a permit, and we collect, clean, and turn the messy public permit, contractor, property, and city council decision data that governments keep but never make usable into the signals our customers make decisions on. We're venture-backed, trusted by some of the largest Fortune 100 companies in the world.
We're hiring a Forward Deployed Engineer: a production engineer who wants to work shoulder-to-shoulder with customers and turn their hardest problems into software.
Customers are where we discover what the platform becomes next, and you sit directly in that loop. You work with a customer to understand a messy problem, explore the data, prototype quickly, prove the solution actually works, then engineer it into something we confidently run in production, working alongside AI agents the way the rest of our engineering team does.
This is not a hand-off role, and it is not a solutions architect or analyst role. You don't gather requirements for engineering, and you don't build a proof of concept that someone else productionizes. You own the technical solution from customer discovery through prototype, rigorous validation, production, and iteration, and you turn what you learn into reusable parts of the Shovels platform.
What you'll do
- Embed with strategic customers: learn their business, data, and workflows deeply enough to find problems they may not know how to articulate yet. Work side by side with Customer Success and Data Insights to join calls, run working sessions, and build the partnership with our top accounts.
- Build production solutions: write and own production code across data pipelines, APIs, applications, agents, and whatever else the problem requires. Some of it lands as hosted feature layers in ArcGIS, some as Snowflake, BigQuery, or Databricks data shares, some as customer-specific agents and skills on the Shovels API and CLI.
- Prototype fast, then prove it works: use notebooks, SQL, scripts, AI agents, or whatever gets you to an answer quickly. Then establish the correctness, edge cases, quality metrics, and failure modes required to put Shovels' name behind it.
- Turn one-offs into product: recognize when a customer-specific solution reveals something the core platform should do, then build it.
- Own the full loop: customer, discovery, build, production, measure, learn, build again. No throwing requirements over the wall, in either direction.
- Stand behind the numbers: if a signal influences a customer's decision, you can explain why we trust it and how we know it works.
- Build with AI agents: like the rest of engineering, you build alongside agents, creating skills, workflows, and tooling that let one strong engineer discover, build, test, and ship what used to take a team.
Who you are
- An engineer first. You've shipped and owned production software. You go through the same interview loop and clear the same bar as the rest of Shovels Engineering.
- You genuinely like customers. Many engineers don't. You do. You get energized by hearing the messy problem firsthand and working out the solution in the room, and you'd take that over a perfectly written spec.
- Strong with Python, SQL, and messy real-world data. You can reason from first principles and sketch a solution without AI as your foundation.
- Living AI engineering, not observing it. You have earned opinions on harnesses, context, and verification from shipped work, and you build with agents by default.
- Comfortable across the stack. The right solution might be a data pipeline, an API, a model, a map, an internal tool, or a customer-facing application. You follow the problem. Geospatial or ArcGIS / Esri experience is a strong plus.
- Rigorous when it matters. You know the difference between "this worked in my notebook" and "we've proven this is something customers can depend on."
- Fast when it doesn't. You know when a day proving an idea beats three weeks engineering the wrong thing.
- Product-minded. You don't just solve the request. You look for the underlying pattern and ask what Shovels should build so the next customer gets it for free.
- Comfortable with ambiguity. There often won't be a spec, because helping figure out what should exist is part of your job.
- Domain-curious. Knowledge of construction, real estate, permitting, or MEP is a plus and can be learned on the job. An appetite to go learn it directly from customers matters more.
You don't need to be the world's best engineer, data scientist, and customer success manager rolled into one. You do need to be a strong engineer who wants to operate at the intersection of customers, data, and product.
If your favorite kind of problem starts with a customer saying "I wonder if we could…" and ends with production software answering "yes, and here's how we know it works," we'd like to talk.
How we work
- U.S.-based, by design for this role: unlike our core engineering roles (which span the Americas and Europe), the Forward Deployed Engineer is U.S.-based and authorized to work in the U.S. You need real-time overlap with U.S. customers and to join their calls live. Engineering is in Europe and business is on the US West Coast, so you're the synchronous side of the company: customer conversations happen in real time, and what you learn goes back to engineering in writing so it survives the 24-hour round trip.
- Radical transparency on financials, strategy, and company direction.
- Weekly AI Lab for trading agent workflows and prompt patterns.
- Build for the customer. A deliverable counts when a customer uses it to make a decision, not when it ships.
- Fully remote, async-friendly, with competitive salary, meaningful equity, and at least twice-yearly in-person gatherings.
How to apply
Send your resume, LinkedIn profile, and a short note on why you're excited about Shovels and this role to petra@shovels.ai. It's a short process, and you hear back either way.
