Software Engineer II, Platform & Agents

About Onshore

Onshore is the modern standard for U.S. companies claiming the R&D tax credit. We find, document, and defend the tax credits and deductions that farms, manufacturers, architects, and engineers have earned. Claiming the R&D tax credit has historically been an ordeal full of consultants, spreadsheets, and hidden fees. Now, our software and AI-powered solutions read every line of a client's records against the tax code, produce verified computations, and are reviewed by a credentialed tax professional before filing.

About the team

The Platform & Agents team builds the infrastructure everyone else at Onshore runs on. This team owns the agent execution framework behind every study, the evaluation harness that proves agent output is correct before an expert sees it, the versioned tax law store that backs our citations, and the tooling operators use to run engagements. When this team ships, the whole company gets faster.

What you'll do

• Build the agent execution framework behind every study: durable multi-step workflows with state that survives crashes, timeouts, and week-long client response gaps

• Extend the evaluation systems that prove agent output is correct before an expert sees it, from golden-set regressions to per-engagement cost and quality observability

• Build the versioned tax law and citation infrastructure that our computation and reporting products consume

• Ship the tooling operators use to run engagements: task management, document collection, and client communication

• Treat other Onshore engineers as your customers, designing primitives and paved paths the Agriculture and Corporate Tax teams build on

• Own what you ship in production, with real SLOs

About you

If you meet the minimum requirements, apply. The preferred qualifications are a bonus, not a gate, and we would rather see your work than a perfectly matched resume.

Minimum requirements

• 2+ years designing, shipping, and operating production software, from requirements through architecture, delivery, and post-launch iteration.

• Strong proficiency in a typed language. We work in TypeScript; deep experience elsewhere with real interest in switching counts.

• High ownership: comfortable with ambiguity, able to define the problem and the plan rather than execute a predefined spec.

• You care about correctness beyond what most product engineering demands. Our output is reviewed by tax professionals and filed with the IRS.

• Clear, precise communication on technical topics, in writing and in person.

• Able to work in person at our New York City office.

Preferred qualifications

• Experience building or operating LLM-powered systems in production: agents, evaluation harnesses, structured extraction, or prompt and skill infrastructure.

• Experience building platforms or internal tools where other engineers were your customers, and judgment about when to build a primitive versus ship a one-off.

• Depth in AWS or comparable cloud infrastructure, especially workflow orchestration, queueing, and relational data at scale.

• Compliance-aware system design: audit trails, data lineage, tenant isolation, or work in regulated domains.

Pay and benefits

The annual base salary for this role is $185,000, with a total equity grant valued at $70,000. Equity is ISOs valued at our Series B preferred price, on a standard four-year vest with a one-year cliff and early exercise available. Benefits include health, dental, and vision coverage and unlimited PTO.

Our interview process

Three conversations, about two hours of live time, with a decision typically one day after each step.

1. Resume deep dive, 30 minutes. We go through your experience and projects at real depth: what you built, the decisions you made, and what you'd do differently. Bring the work you're proud of.

2. Take-home assessment, then a one-hour review. You build something on your own time, then walk us through it live: your assumptions, patterns, and tradeoffs. We care as much about how you reason through the decisions as the code itself.

3. Behavioral interview, 30 minutes. How you work, how you handle ambiguity and disagreement, and whether Onshore is the right place for you, in both directions.

No six-round gauntlets, no trick questions, no waiting weeks to hear back.