We're building an AI-powered insurance brokerage that's transforming the $900 billion commercial insurance market by automating processes that currently run on pre-internet systems. Fresh off our $8M seed round, we're looking for an exceptional Applied AI Engineer who can architect and develop intelligent agent systems across our entire customer journey.
You'll create sophisticated AI systems that turn traditional insurance processes into efficient, data-driven experiences powered by both ambient and frontier AI agents. Ambient agents will work continuously in the background, hooking into context and memory, while frontier agents will serve as the critical interface between ambient agents and human operators - delivering low-latency, high-quality collaboration experiences. This role sits at the critical intersection of engineering, AI, and business operations—a position to directly shape how we scale from seed to industry leader.
We're committed to "Staying REAL" with our AI systems - building agents that are Reliable, Experience-focused, Accurate, and have Low latency. You will work directly with the CEO and CTO to execute on our AI vision with a bias toward action. We live by core principles: "There is no try, there is just do," "Actions lead to information, always default to action," and "Strong opinions lead to information." We need engineers who build and ship, not just plan and strategize.
Design and implement AI agent architectures using temporal.io workflows and pydantic-ai, adhering to distributed systems best practices
Develop deep understanding of CAP theorem tradeoffs between Consistency, Availability, and Partition tolerance to build resilient agent systems
Implement Lambda architecture patterns combining event streaming for real-time processing with batch processing for comprehensive analytics
Design and develop frontier agents that provide exceptional human-AI collaboration interfaces with low latency and high accuracy
Create ambient agents working in background processes that connect seamlessly to frontier agents for a unified system
Build sophisticated voice AI systems for frontier agents that facilitate natural, contextually rich conversations with customers
Build highly Reliable ambient agents that monitor event streams and proactively take action based on triggers, with robust recovery mechanisms and fault tolerance
Architect communication protocols between ambient agents and frontier agents to ensure seamless information flow
Build retrievers and RAG systems that instantly surface relevant policy information and domain knowledge with high Accuracy during interactions
Develop self-service flows for customer onboarding and policy management that maximize conversion and deliver superior user Experience
Architect human-in-the-loop workflows that smoothly transition between AI and human experts with minimal Latency
Design comprehensive observability with logfire to monitor agent Reliability, performance, and outcomes
Establish rigorous evaluation frameworks to continuously improve agent Accuracy and optimize conversion rates
You're experienced with our REAL framework for building reliable, experience-focused, accurate, and low-latency AI systems
You have deep expertise with distributed systems, event sourcing, and scalable architecture patterns
You understand CAP theorem tradeoffs and can make appropriate architectural decisions for different system requirements
You can implement Lambda architecture combining real-time event processing with batch analytics pipelines
You're equally comfortable with TypeScript/Node.js and Python/ML frameworks
You're proficient with modern agent orchestration frameworks and temporal.io workflows
You ship features daily and take immediate action instead of overthinking
You embrace "there is no try, there is just do" as your engineering mantra
You understand that actions lead to information, and default to shipping code
You hold strong opinions but remain open to learning from real-world results
You're a pro at using AI coding assistants like Cursor or WindSurf to accelerate development
You excel at delegating logic to AI to achieve outcomes while maintaining oversight
You thrive at the intersection of engineering, data, AI, and business strategy
Strong experience with both TypeScript/Node.js and Python
Experience with temporal.io workflows and pydantic-ai for building durable agent systems
Deep understanding of distributed systems principles, CAP theorem tradeoffs, and event sourcing architecture
Experience designing systems that balance consistency, availability and partition tolerance based on specific use cases
Proficiency in implementing Lambda architecture with both real-time event streaming and batch processing
Proven track record building sophisticated voice AI or multi-channel communication systems
Experience with multi-channel communication platforms (voice, email, chat, web)
Strong frontend development skills with Next.js/React
Proven ability to implement systems that directly improved business metrics
Advanced usage of Cursor or WindSurf coding IDE
Must be based in San Francisco and work in-office 5.5 days per week (relocation assistance provided) We're building a modern, AI-native infrastructure to power our growth:
Temporal.io for durable workflow orchestration across agent systems (for Reliability)
Pydantic-AI for type-safe agent development with structured validation (for Accuracy)
Event sourcing architecture with Redis streams and PostgreSQL (for Reliability)
Lambda architecture combining real-time event streams and batch processing
RAG systems with rigorous evaluation frameworks for domain-specific knowledge (for Accuracy)
Optimized response processing for minimal user waiting times (for Latency)
Claude (Anthropic), GPT-4.1 (OpenAI), and select open source models
Model Context Protocol (MCP) for optimized LLM interactions (for Latency)
Logfire for comprehensive agent observability and analytics (for the entire REAL framework)
User-centered design principles for intuitive agent interactions (for Experience)
LiveKit for real-time, bidirectional voice communication
Twilio for enterprise-grade telephony infrastructure
OpenPhone for capturing and analyzing human conversations
TypeScript/Node.js for robust application development
Python for AI systems and ML workflows
Next.js/React for frontend experiences and customer portals
Temporal workers for distributed, fault-tolerant process execution
Event-driven architecture with Redis streams for scalable applications
Event sourcing with PostgreSQL for transactional data
Redis for high-performance caching and messaging
Vector databases for semantic search capabilities
Build and deploy outreach agent underwriter submission load balancer that manages quoting across human underwriters and online portals
Design and implement the core architecture for ambient agents using temporal.io and pydantic-ai
Create voice AI system capable of handling basic insurance inquiries with low latency and high reliability
Set up comprehensive observability with logfire to implement the REAL framework metrics
Create dashboards that provide real-time visibility into user experience quality and agent accuracy
Develop graphs of agents via temporal and pydantic-ai to handle complex insurance workflows
Build customer service agents specialized for policy renewals and claims processing
Design intelligent escalation paths that seamlessly transition from AI to human experts
Implement AI forms and voice AI to collect information via product-led growth channels
Create outreach AI agents that nurture leads through personalized outbound campaigns
Develop autonomous feedback loops where conversion and satisfaction data improves targeting
Build personalization engines that adapt messaging based on prospect characteristics
Implement comprehensive analytics that identify conversion bottlenecks
Scale AI agents for end-to-end automation of key insurance processes
Integrate all agent systems into a unified, observable architecture with logfire
Staying REAL: Our framework for building AI agents that are:
Reliable: Agents that consistently perform, handle edge cases, and recover gracefully from failures
Experience-focused: Creating intuitive, natural interfaces for human-AI interaction
Accurate: Ensuring high-quality outputs through rigorous evaluation frameworks and feedback loops
Low latency: Delivering smooth, responsive interactions that feel conversational and immediate
Ambient + Frontier Agents Architecture: Build a two-tier agent system where ambient agents work continuously in the background handling routine tasks, while frontier agents provide the human-AI collaboration layer with low-latency, excellent experience and accurate communication between systems
Event-Driven Architecture: Design systems that react to events and state changes for maximum responsiveness
Distributed & Durable: Create fault-tolerant agents that maintain state and recover from failures
CAP Theorem Understanding: Make intelligent tradeoffs between consistency, availability, and partition tolerance based on specific use cases
Lambda Architecture Approach: Combine event streaming for real-time processing with batch processing for complete analytics
Action Orientation: Always default to action - ship code, gather data, and iterate rather than overthink or overplan
Execution Focus: There is no try, there is just do - we value engineers who build and ship, not just plan and strategize
Strong Opinions: Form and express clear viewpoints that can be tested against reality to generate valuable information
Revenue Focus: Tie every initiative directly to business outcomes and improved metrics This is an early-stage role at a fast-moving startup, and you'll often experience the crawl-walk-run approach to building. You'll quickly prototype applications and then push them into productionized systems that can scale. We're looking for people who can be creative in providing impact first, then take learnings from that impact and push them back into the system.
You should ideally have worked in an early-stage startup environment and understand the pacing. This is a fast-paced environment where we value ownership and quick, rapid feedback loops within the team. You'll work directly with the CEO and CTO to execute on our AI vision with a bias toward action.
We require you to be in San Francisco and work from our office 5.5 days per week. We'll cover relocation costs and believe the best teams collaborate intensively in person.
TypeScript, Node.js, Python, Temporal.io, Pydantic-AI, Event Sourcing, Distributed Systems, CAP Theorem, Lambda Architecture, Next.js/React, Model Context Protocol, RAG Systems, Voice AI, Vector Databases, Redis Streams, PostgreSQL, Full-Stack Development
Full-time
$118K–$160K
San Francisco, California