Agentic Product Builder

Gravity; United States; Remote; USD 140,000 - 230,000; Full time

Benefits offered by this job

Work from anywhere

Job summary

Gravity is seeking a product-minded AI agent engineer to own AI features from problem through implementation, evaluation and release. You will shape workflows for a public-facing product, build them yourself, and improve them using evidence from real usage.

Responsibilities include breaking user needs into agent tasks, writing requirements and test scenarios, creating workflows with LangGraph/LangChain, designing prompts and memory, and evaluating system quality.

Qualifications

• Experience shipping real products in product, engineering or similar role.
• Practical LangGraph or LangChain work with demonstrable repo, demo or live product.
• Understanding of tool use, planning, state, multi-step workflows, memory and context limits in agent systems.
• Ability to assess quality and cost using tracing, evaluation data, token usage and latency.
• Comfortable working in Python and turning ambiguous problems into clear plans.
• Evidence of building, shipping and repairing agents in production or real use.

Responsibilities

• Break user needs into tasks that agents can handle, defining boundaries and when people should stay involved.
• Write concise requirements, success measures and test scenarios before building.
• Create agent workflows with LangGraph and LangChain, connecting them to relevant tools, APIs and data.
• Design prompts, context and memory, then evaluate quality and refine the system.
• Investigate traces and failures such as looping, incorrect tool use, hallucinations and context limits; manage latency and operating cost.
• Ship iteratively, document the work, and use user feedback to choose what to improve next.

Skills

Problem decomposition

Cross-functional collaboration

Test planning

Performance optimization

Cost awareness

Tools

LangGraph

LangChain

Python

APIs

Data integration

Vector databases

Retrieval-augmented generation

Job description

Role overview

Own AI product features from the initial problem through implementation, evaluation and release. This hands-on role combines product judgment with agent engineering: you’ll shape workflows for a public-facing product, build them yourself, and improve them using evidence from real usage.

Responsibilities

• Break user needs into tasks that agents can handle, defining boundaries and when people should stay involved.
• Write concise requirements, success measures and test scenarios before building.
• Create agent workflows with LangGraph and LangChain, connecting them to relevant tools, APIs and data.
• Design prompts, context and memory, then evaluate quality and refine the system.
• Investigate traces and failures such as looping, incorrect tool use, hallucinations and context limits; manage latency and operating cost.
• Ship iteratively, document the work, and use user feedback to choose what to improve next.

Requirements

• Experience shipping real products in a product, engineering or similar role; prototypes alone are not enough.
• Practical LangGraph or LangChain work that can be demonstrated through a repository, demo or live product.
• Understanding of tool use, planning, state, multi-step workflows, memory and context limits in agent systems.
• Ability to assess quality and cost using tracing, evaluation data, token usage and latency.
• Comfortable working in Python and turning ambiguous problems into clear plans.
• Evidence of building, shipping and repairing agents in production or real use.

Nice to have

• Production experience with retrieval-augmented generation, embeddings or vector databases.
• Familiarity with structured outputs, function calling, guardrails, multiple model providers, multi-agent patterns or MCP.
• Experience deploying and monitoring AI features, including APIs, queues, observability and cloud infrastructure.

Benefits and work setup

• Fixed pay plus performance-based compensation tied to revenue generated by shipped agents.
• Direct access to product decisions and close collaboration with the founder.
• Work from anywhere.