Staff Software Engineer

PeopleLoop

About The Position

PeopleLoop is an AI-powered experience that modernizes benefits and PEO solutions for people-driven companies. It reduces complexity and errors across critical people operations, gives leaders clearer insight into compensation and benefits costs, trends, and tradeoffs, and delivers a more modern employee experience. In this role, you will lead the design and delivery of the core services that every PeopleLoop product depends on—including events, audit, EDI, the compliance graph, and the enrollment orchestrator. You’ll also help build the platform our AI agents run on: tool contracts, retrieval over the compliance graph, LangGraph‑style orchestration, and guardrails. As a Staff‑level engineer, you’ll set patterns, mentor engineers, drive design reviews and RFCs, and own the quality bar for critical platform systems. Most services are built in Python (FastAPI), with some in Node/TypeScript (NestJS).

Requirements

7+ years of software engineering experience, with 2+ years at Senior or Staff level owning cross‑team initiatives
Strong production experience building Python services; working proficiency in Node/TypeScript
Deep experience with MongoSQL, PostgreSQL and at least one of Kafka/MSK, Kinesis, or EventBridge
Hands‑on AWS experience, including Step Functions or a comparable workflow engine
Shipped at least one LLM‑ or agent‑backed feature to production, including retrieval, tool use, and guardrails (LangGraph, LangChain, LlamaIndex, or similar)
Experience designing and operating multi‑tenant systems with strong security and privacy guarantees
Comfortable writing RFCs, mentoring engineers, and driving designs from concept through production
Bachelor’s degree required, preferably in Computer Science, Engineering, or a related technical discipline; equivalent experience will be considered.

Nice To Haves

Benefits, PEO, or EOR domain experience (834, 820, QLE, COBRA, multi‑country payroll or employment)
Production use of graph databases (Neptune, Neo4j) with Cypher or Gremlin
Workflow engines such as Temporal, Step Functions, or Airflow
Authorization engines and policy‑as‑code (Cedar, OPA)
CDC pipelines (Debezium, DMS) and lakehouse architectures
DSLs or rules engines for carrier mappings or compliance checks
Agent evaluation tooling (LangSmith, Ragas, Braintrust) and prompt or model registries
Vector stores and hybrid retrieval (pgvector, OpenSearch, Pinecone)
MCP servers, function‑calling contracts, or structured‑output patterns

Responsibilities

Define clear system boundaries and ownership across Postgres (system of record), the compliance graph, Python (FastAPI) and Node/TypeScript (NestJS) services, event pipelines, AI agents, and external partners (carriers, EDI, payroll).
Build and maintain an append‑only, tamper‑evident audit log across services. Provide query APIs for compliance and support use cases, with retention and legal‑hold support. Ensure every AI agent action is audited end‑to‑end.
Own 834 eligibility and 820 premium flows, including partner variants. Build carrier adapters with strong contract tests, reconciliation, and replay so new integrations are primarily configuration, not bespoke engineering work.
Model jurisdictions, regulations, plans, and eligibility rules in Neptune (or Neo4j). Own graph modeling, bulk loading, query design, and safe retrieval APIs that services and AI agents use to answer “what applies here?”
Design and evolve long‑running workflows (Step Functions or Temporal) for open enrollment, QLEs, terminations, and carrier submission. Support agent‑assisted steps, simulation vs. live runs, human approvals, and reliable replay and recovery.
Help build the foundational agent platform: tool contracts over internal APIs and the compliance graph, retrieval and grounding, LangGraph or similar orchestration, prompt and model registries, evaluation harnesses, cost and latency controls, and PII‑safe logging.
Design and maintain strong tenant isolation (RLS, per‑tenant keys), API gateway and authorization (Cedar), secrets and KMS, rate limits, and quotas—including per‑tenant agent budgets.
Deliver meaningful observability with OpenTelemetry traces, metrics, and logs across services and agent runs. Define SLOs that drive action. Own CDC from Postgres into the lake for analytics and agent evaluation datasets.
Improve developer velocity through shared Python and TypeScript libraries, OpenAPI/GraphQL code generation, and tests that cover workflows, integrations, graph queries, and agent behavior.