View job listing
Apply to Staff Software Engineer - AI Engineering
Staff Software Engineer, AI Engineering
What they're building
A newAI-native billing platform built from scratch on GCP. It replaces the manual work of US medical billing (claims, eligibility checks, prior authorization, remittances) with LLMs, orchestrated agent workflows and RAG
Stack: Python, LangGraph, LangChain
· GCP, Vertex AI · Kafka, Pub/Sub · Java, Spring Boot
· pgvector, Pinecone, ChromaDB
· Docker, Kubernetes, FastAPI.
The role
Lead the hands-on implementation, design and adoption ofAI agents across the billing and revenue-cycle platform. A senior IC who influences technical strategy across teams and sets standards for reliable, secure AI systems.
Own hands-on implementation of production AI agents and agentic workflows, mainly in Python (LangChain, LangGraph).
Design, test and operate AI services that combine agent orchestration, APIs, retrieval
and models for real customer use.
Connect Python agents to Java/Spring Boot platform services through APIs, events and contracts; contribute in both.
Integrate LLMs, ML models and enterprise data with strong patterns for reliability, observability and security.
Build bounded, auditable AI workflows with fallbacks, evaluation, authorization boundaries and human review.
Turn workflow pain points into software with measurable impact, alongside product,
design, data and ops.
Review code across AI services and Java codebases; mentor junior and mid-level
engineers, including Java-first ones.
Improve latency, throughput, quality and cost across agent workflows and backend
services.
Language bar: expert in Python or Java, working proficiency in the other. Python-first with real Java comfort is ideal. Java-first with deep Python AI work and hands-on LangGraph also works.
Requirements
8+ years building scalable distributed systems.
2–3+ years designing, shipping and maintaining production AI applications that combine LLMs, application logic and business workflows.
Hands-on with LangGraph, LangChain, LlamaIndex or CrewAI; deep understanding of RAG, tool calling, state management and human-in-the-loop.
Distributed and event-driven systems with Kafka or similar.
Multi-step agents with tools, retrieval, memory, structured outputs and explicit control flow, and the ability to explain the trade-offs.
Production AI safeguards: observability, latency and cost monitoring, automated evaluation, fallbacks.
Cloud-native on AWS, Azure or GCP.
Preferred: healthcare IT, RCM, billing, claims, fintech or another regulated domain (the HM estimates a 6–18 month ramp without it) · autonomous multi-agent systems with multi-tier memory
· knowledge graphs or enterprise vector search
· PyTorch, HuggingFace, embeddings or fine-tuning
· founding engineer or principal architect who drove org-wide AI adoption.
What wins here
LangGraph or LangChain named in the role descriptions, not just the skills list.
Multi-agent orchestration in production: coordinator and sub-agents, plan-andexecute patterns.
Stateful long-running workflows with retries, failure recovery and partial resumption.
RAG ownership with detail: chunking strategy, embeddings, hybrid retrieval,
evaluation.
LLM evaluation method: golden datasets, LLM-as-judge, RAGAS, Langfuse, Arize.
This was the most common gap in our screens.
Kafka or Pub/Sub in production, GCP or Vertex AI, MCP in production, AI coding tools
in daily use.
What they're building
A newAI-native billing platform built from scratch on GCP. It replaces the manual work of US medical billing (claims, eligibility checks, prior authorization, remittances) with LLMs, orchestrated agent workflows and RAG
Stack: Python, LangGraph, LangChain
· GCP, Vertex AI · Kafka, Pub/Sub · Java, Spring Boot
· pgvector, Pinecone, ChromaDB
· Docker, Kubernetes, FastAPI.
The role
Lead the hands-on implementation, design and adoption ofAI agents across the billing and revenue-cycle platform. A senior IC who influences technical strategy across teams and sets standards for reliable, secure AI systems.
Own hands-on implementation of production AI agents and agentic workflows, mainly in Python (LangChain, LangGraph).
Design, test and operate AI services that combine agent orchestration, APIs, retrieval
and models for real customer use.
Connect Python agents to Java/Spring Boot platform services through APIs, events and contracts; contribute in both.
Integrate LLMs, ML models and enterprise data with strong patterns for reliability, observability and security.
Build bounded, auditable AI workflows with fallbacks, evaluation, authorization boundaries and human review.
Turn workflow pain points into software with measurable impact, alongside product,
design, data and ops.
Review code across AI services and Java codebases; mentor junior and mid-level
engineers, including Java-first ones.
Improve latency, throughput, quality and cost across agent workflows and backend
services.
Language bar: expert in Python or Java, working proficiency in the other. Python-first with real Java comfort is ideal. Java-first with deep Python AI work and hands-on LangGraph also works.
Requirements
8+ years building scalable distributed systems.
2–3+ years designing, shipping and maintaining production AI applications that combine LLMs, application logic and business workflows.
Hands-on with LangGraph, LangChain, LlamaIndex or CrewAI; deep understanding of RAG, tool calling, state management and human-in-the-loop.
Distributed and event-driven systems with Kafka or similar.
Multi-step agents with tools, retrieval, memory, structured outputs and explicit control flow, and the ability to explain the trade-offs.
Production AI safeguards: observability, latency and cost monitoring, automated evaluation, fallbacks.
Cloud-native on AWS, Azure or GCP.
Preferred: healthcare IT, RCM, billing, claims, fintech or another regulated domain (the HM estimates a 6–18 month ramp without it) · autonomous multi-agent systems with multi-tier memory
· knowledge graphs or enterprise vector search
· PyTorch, HuggingFace, embeddings or fine-tuning
· founding engineer or principal architect who drove org-wide AI adoption.
What wins here
LangGraph or LangChain named in the role descriptions, not just the skills list.
Multi-agent orchestration in production: coordinator and sub-agents, plan-andexecute patterns.
Stateful long-running workflows with retries, failure recovery and partial resumption.
RAG ownership with detail: chunking strategy, embeddings, hybrid retrieval,
evaluation.
LLM evaluation method: golden datasets, LLM-as-judge, RAGAS, Langfuse, Arize.
This was the most common gap in our screens.
Kafka or Pub/Sub in production, GCP or Vertex AI, MCP in production, AI coding tools
in daily use.