Forward Deployed Engineer

About the role

You will act as the primary technical translator between business stakeholders and engineering execution, owning the full lifecycle of a Claude-powered internal tool from initial discovery through production deployment. You will partner deeply with at least one business function, treating their workflow as a product and their data as a core dependency that you are responsible for understanding and shaping. Your day-to-day will involve writing, testing, and personally maintaining code in production, where you are the single point of accountability for reliability, cost, and user adoption of the systems you ship. You will define and defend the engineering standards that determine whether an agent is a fragile demo or a durable, unattended service that the business can rely on. This role requires a builder who is comfortable making technical tradeoffs without a product manager in the room and who measures success through the impact of their automation on real workflows. You will be expected to evangelize best practices, turning a successful point solution into a repeatable pattern that other teams can adopt and extend. Ultimately, you will drive a step change in operational efficiency by replacing manual, repetitive work with intelligent, self-running agents that create tangible time and cost savings for the company.

Key facts

What you'll do

• Discover and scope ambiguous business problems by running structured technical interviews with stakeholders to translate vague ideas into a well-defined build plan.
• Build end-to-end Claude-powered agents and automations in production using Claude Code, the Claude API, or agentic frameworks, taking a feature from zero to a live service with direct oversight.
• Engineer for long-term reliability by making deliberate choices about model selection, context management, caching strategies, and error handling to ensure robustness beyond a minimal viable demo.
• Establish and monitor key performance indicators before writing any code, defining the metric that proves impact on time saved, error reduction, or throughput gain.
• Enable business teams by training them to construct their own lightweight automations, reducing future dependency on the centralized AI engineering role.
• Influence cross-functional stakeholders by translating complex technical constraints into clear business tradeoffs, moving discussions from skepticism to committed adoption.
• Create reusable patterns and documentation that allow other builders within Taskrabbit to replicate successful automation architectures across different functions.
• Maintain and iterate on deployed services to ensure they continue to perform as data volumes, user behavior, and business requirements evolve over time.
• Prioritize ruthlessly on cost and latency, ensuring that the agent's operational footprint remains efficient and justifiable at scale.
• Act as the on-call expert for the systems you ship, diagnosing issues in production and applying fixes without disrupting the broader business operations.

Requirements

• 2+ years of hands-on experience building AI agents, automations, or internal tools professionally, with direct, independent production deployment of Claude-powered systems.
• Demonstrated ability to write, test, and ship production code that calls an LLM as part of an agent or data pipeline, with a portfolio of deployed projects you can discuss in technical detail.
• Strong software engineering fundamentals, including proficiency in at least one general-purpose programming language, data structures, and distributed system concepts relevant to agentic workflows.
• Fluency in discussing engineering decisions such as context window management, token efficiency, model selection, and failure modes, not just workflow configuration.
• Experience with measurement and instrumentation, having set up logging, tracing, and evaluation frameworks to validate the impact of automated systems.
• Comfort operating in an ambiguous, fast-paced environment where requirements evolve and you must balance speed with long-term maintainability.
• A pragmatic mindset that prioritizes simple, robust solutions over clever but fragile architectures, especially when operating systems that touch real business data.
• Willingness to work a hybrid schedule with two mandatory in-office days on Tuesdays and Wednesdays at the San Francisco location on 130 Sutter Street.

Nice to have

• Experience with Claude's agentic tool use, function calling, and prompt engineering patterns specific to production systems.
• Familiarity with internal developer platforms or infrastructure that can host and monitor long-running agent workflows.
• Background in operations or DevOps practices that help you own reliability, monitoring, and cost management for AI services.
• Previous work in a marketplace or high-volume transactional environment where automation directly impacts revenue or customer experience.

Practical notes

This role requires two days per week in the office on Tuesdays and Wednesdays at the San Francisco office located at 130 Sutter Street. The position is full-time and eligible for Taskrabbit's comprehensive benefits package and equity plan. Relocation assistance may be available for qualified candidates.