Senior Software Engineer - Core AI

Navan•Palo Alto, CA

About The Position

We're seeking a Senior Software Engineer to build our Composable AI Microservice Architecture—a fleet of hyper-specialized AI services. You will design and develop intelligent, production-grade agents that transform raw LLMs into task-specific services with high agency and deterministic precision. This is a high-velocity building role where you'll ship production code that augments high-leverage human workflows across the company.

Requirements

Product-minded AI building experience. You are a Senior or Staff-level engineer who loves building features that people actually use, understanding that valuable AI solves specific human workflow problems.
Data-first engineering mindset. You are skilled at data engineering and serialization, knowing that a well-formatted context and the right data are more valuable than endless prompt tweaking.
Agentic systems specialization. You have experience building agentic flows using frameworks like AWS Agent Core, Google Vertex ADK, or custom autonomous agent runtimes, and know how to balance agency with precision.
Experimentation and evaluation expertise. You trust benchmarks over intuition and are comfortable building observability pipelines and writing custom evals to measure agent performance, reliability, and success metrics.
High-velocity shipping ability. You thrive in production environments where success is measured by the number of high-leverage human workflows you successfully augment, not research papers published.

Responsibilities

Develop specialized AI microservices.
Design and build the core logic for task-specific AI agents using AWS Bedrock Agents, Google ADK, or Anthropic Claude SDK, owning the full implementation from data serialization, memory to agentic tool-use, MCP (Model Context Protocol) gateway integrations, and multi-step reasoning.
Drive optimization-first development.
Improve service quality through a rigorous hierarchy: context engineering and skills scripting for optimal knowledge delivery, logic and tooling design for real-world interaction, and prompt and behavioral tuning for specific, repeatable behaviors.
Build for the fleet.
Ensure every service you build is "Core AI Ready" by fully integrating with observability and evals frameworks like LangSmith, Braintrust for agent tracing, drift detection, versioning, and success metrics.
Lead systemic experimentation.
Design and run controlled experiments across different models and datasets using evals frameworks and benchmarking tools to prove that your improvements are attributable and reproducible.
Collaborate on architecture.
Work closely with AI Ops to ensure your services are deployable at scale using autonomous agent runtimes, agent sandboxing (E2B, AWS Lambda), and AWS AgentCore infrastructure, and with the Platform team to deliver "Demo to Prod" leverage.