Forward Deployed Engineer

This is a hybrid role that will require two days in-office each week on Tuesdays and Wednesdays at our SF location on 130 Sutter Street. 

About the Role

Every business function at Taskrabbit — Marketing, Customer Support, Finance, Operations — is a "customer" with real workflows, real data, and real friction. Your job is to embed with them, scope their use cases, and build the Claude-powered agent, automation, or tool that solves them.

This is an internal-facing role — there is no external customer or product work. Reporting to the Director of AI Strategy and Enablement, you'll operate the way an FDE operates at a high-growth AI company: full ownership of a deployment from discovery through production and direct accountability for whether what you ship actually changes a metric. In most cases you'll own a build end-to-end solo; in some functions you may partner with that team's own subject-matter expert to pair domain depth with technical execution.

We are hiring for engineering depth first. This is not a role for someone whose experience is limited to no-code/low-code automation platforms (Zapier, Make, n8n) without underlying software engineering fundamentals. The bar is: you have personally written, tested, and shipped production code that calls an LLM — ideally Claude — as part of an agent, pipeline, or internal tool, and you can speak fluently to the engineering decisions (not just the workflow decisions) behind it.

What You'll Do

• Discover & scope: Embed with a business function, run structured technical discovery, and turn an ambiguous use case idea  into a scoped build. Build in production: Independently design, write, test, and deploy Claude-powered agents, automations, and internal tools (Claude Code, Claude Cowork, Claude API/Agent SDK, or agentic frameworks built on Claude) — owning the full lifecycle from first commit to something running unattended in production.
• Engineer for reliability, not demos: Make and defend real engineering decisions — model selection, context/caching strategy, evals, human-in-the-loop checkpoints, error handling, cost forecasting — the difference between a working prototype and something that survives contact with real data and real users.
• Measure: Establish a baseline before every build; define the metric that proves impact (time saved, error rate, throughput) and instrument for it from day one.
• Enable: Leave each function more capable than you found it — train non-technical teams to identify and eventually build their own lightweight automations, without creating dependency on this role.
• Influence: Translate between business problems and technical tradeoffs; move skeptical stakeholders from "prove it" to adoption; align priorities with senior leaders and company OKRs.
• Scale patterns: Turn what works in one function into a reusable playbook other teams — and other builders — can pick up.

What We're Looking For

• Production Claude experience (required): 2+ years hands-on building AI agents, automations, or internal tools professionally, with direct, independent, code-level work in Claude Code, the Claude API, Claude Skills/Agent SDK, or an agentic framework built on top of Claude. You should be able to walk us through a system you built, why you made the architecture choices you did, and what broke in production.
• Software engineering fundamentals: Comfortable writing and owning real code — not exclusively configuring no-code/automation platforms. You should be fluent in things like API integration, MCP, version control, testing, and debugging a system you didn't just prototype but actually run.
• Technical judgment: You surface model selection tradeoffs, caching and cost strategy, reliability/failure modes, and human-in-the-loop design without being asked — because you've had to make those calls before and lived with the consequences. You also know when a workflow's data sensitivity should change the build: what a system can safely touch, and what needs scoping, redaction, or a human review step before it does.
• Business partnership: Track record embedding with non-technical teams, running discovery, and converting workflows into scoped builds — including the judgment to say no to AI when it's not the right fix.
• Enablement mindset: You've built capability in other people, not just built things for them.
• High empathy: Adept at sitting with a skeptical or overwhelmed team member long enough to actually understand their workflows, not just extract requirements
• Stakeholder influence: Comfortable presenting to and moving senior leaders and frontline teams alike.
• Impact measurement: Data-driven by default — you define baselines and metrics from scratch and translate results into business terms.
• Builder mindset: Scrappy, self-directed, comfortable operating without dedicated engineering support in a fast-paced, resource-constrained environment.
• Industry (preferred): Marketplace or gig-economy company (e.g., Airbnb, Lyft, DoorDash) or high-growth startup.

Compensation & Benefits: 

At Taskrabbit, our approach to compensation is designed to be competitive, transparent, and equitable. Total compensation consists of base pay + bonus + benefits + perks. The base pay range for this position is $175,000 - $225,000. This range is representative of base pay only, and does not include any other total cash compensation amounts, such as company bonus or benefits. Final offer amounts may vary from the amounts listed above and will be determined by factors including, but not limited to, relevant experience, qualifications, geography, and level.