Software Engineer (AI)

Software Engineer (AI)

The Team + The Role

Pendo spent a decade helping companies understand how people use their software. Agents are the next surface, and this team builds the product for it: customers install Pendo on their own agentic systems, including MCP servers, agent frameworks, and copilots, and Pendo shows them what those agents are actually doing. The team helps customers understand which tools get called, where calls fail, where a contract confuses the model, and what to fix next.

As an AI Software Engineer, you will own features end to end across the service, interface, and the seam between them. You will work across Go services, Vue interfaces, install experiences, instrumentation, telemetry, and LLM-backed product capabilities. You will start with well-shaped problems and take on progressively more ambiguous work as you build context, writing code others can operate and raising issues early when a spec does not match reality.

This role is based in our New York City office.

What this looks like day-to-day

Full-stack feature delivery: Ship full-stack features end to end across Go services and Vue interfaces. You will work across backend and frontend boundaries to build product experiences that are reliable, usable, and ready for production.

Customer install and instrumentation: Build the install and instrumentation experience customers drop into their own agentic systems. Your work should be quick to adopt, hard to get wrong, and light enough to run in production.

Agent behavior visibility: Capture agent behavior and make it legible for customers. You will help expose tool calls, failures, latency, and the places where a model and a tool contract talk past each other.

LLM-backed product development: Work on LLM-backed features including prompt and context design, tool calling, evaluation, and related failure modes. You will use hands-on judgment from building with models to improve how customers debug and understand agent behavior.

Safe, incremental shipping: Ship in small increments behind feature flags and test to the level of product and technical risk. You will use tools such as Jest and Cypress while accounting for non-deterministic model output.

Production ownership: Own what you ship in production through logging, monitoring, fast resolution, and follow-up work that reduces recurrence. You will write code others can operate and support.

AI-enabled workflow: Use AI tools as a normal part of your development workflow while staying accountable for what ships. You will bring curiosity and discipline to how AI improves your speed, quality, and judgment.