Staff Software Engineer, AI & App Experience

Chime Chime · Fintech · San Francisco, CA · AI & App Experience Engineering

Staff Software Engineer to lead technical direction and architecture for AI-powered financial assistant (Jade). Role involves designing, building, and scaling member-facing AI capabilities, including agent systems, evaluation frameworks, and guardrails. Focus on shipping production code, improving reliability and performance, and championing AI-native development practices at scale.

What you'd actually do

  1. Set the technical direction and architecture for how Chime builds with LLMs on Jade: the agent architectures, prompt strategies, and orchestration patterns that shape how Jade reasons and acts, and that other engineers build on
  2. Design, build, and scale new member-facing capabilities for Jade, from prototype through production, moving fluidly between product discovery and hands-on engineering
  3. Build the eval frameworks, observability, and guardrail systems that let the team ship LLM-powered features with speed, safety, and confidence
  4. Develop and harden the backend services and internal tooling behind Jade (model routing, prompt management, agent orchestration, and evaluation pipelines), improving reliability and performance as we scale
  5. Leverage AI and LLMs natively in your own workflow, using AI-assisted coding and rapid prototyping, and turn one-off AI workflows into reusable systems (agent loops, evals, custom tooling) that compound the whole team's output

Skills

Required

  • 8+ years of backend or full-stack software development experience in production environments
  • Deep expertise in system design, distributed systems, and architectural patterns for high-scale systems
  • Proficiency with Python or comparable frameworks, with the breadth to make sound decisions across the stack
  • AI-native fluency: you actively build with LLMs, AI code assistants, and generative AI tooling as a daily part of your workflow, not as a side project
  • Hands-on experience building or shipping LLM-powered product features (agents, conversation experiences, evals, prompt strategies, or guardrails) at production scale
  • A track record of turning AI into durable leverage: reusing and improving workflows, encoding recurring fixes into evals, rules, and tooling instead of solving the same problem twice, and acting as the first and most critical reviewer of AI-generated output
  • Sound judgment about where and how to apply AI, calibrating verification effort and model autonomy to the stakes, reversibility, and cost of each task
  • Experience with transactional databases (e.g., Postgres) and caching systems (e.g., Redis), and a strong focus on writing maintainable, well-tested code
  • Demonstrated technical leadership across teams: setting direction, driving architectural decisions, and aligning cross-functional stakeholders without relying on positional authority
  • A track record of mentoring engineers and raising the technical bar
  • A product mindset with a bias toward action: you ship v1 fast, learn from real usage, and iterate, letting data settle debates

Nice to have

  • experience with agent development
  • fintech
  • startup environments

What the JD emphasized

  • architecting the agent systems, evaluation frameworks, and guardrails that let us ship an AI product reliably and safely at the scale of millions of members
  • shipping production code
  • building the eval frameworks, observability, and guardrail systems that let the team ship LLM-powered features with speed, safety, and confidence
  • develop and harden the backend services and internal tooling behind Jade
  • AI-native fluency: you actively build with LLMs, AI code assistants, and generative AI tooling as a daily part of your workflow, not as a side project
  • Hands-on experience building or shipping LLM-powered product features (agents, conversation experiences, evals, prompt strategies, or guardrails) at production scale
  • A track record of turning AI into durable leverage: reusing and improving workflows, encoding recurring fixes into evals, rules, and tooling instead of solving the same problem twice, and acting as the first and most critical reviewer of AI-generated output
  • Sound judgment about where and how to apply AI, calibrating verification effort and model autonomy to the stakes, reversibility, and cost of each task

Other signals

  • building LLM-powered products at scale
  • shipping AI product reliably and safely
  • architecting agent systems, evaluation frameworks, and guardrails