Design and ship multi-step agentic systems for KYB, underwriting, case review, and risk monitoring workflows.
Architect agent graphs in LangGraph or comparable frameworks with explicit state, durable execution, retries, and safe fallbacks.
Build and harden retrieval layers using chunking strategies, hybrid search, reranking, and grounded citation across financial and business data.
Own evaluation systems including golden sets, offline regression suites, LLM-as-judge, online A/B and shadow evaluations, and red-teaming.
Integrate agents into production systems through typed tools, MCP servers, and existing decisioning, case management, and crosswalking services.
Drive production MLOps including deployment, versioning, traffic shaping, cost and latency budgets, observability, and incident playbooks.
Partner with security, compliance, and legal to maintain SOC 2, GDPR, CCPA, and fair-lending posture.
Translate ambiguous product ideas into agent designs, prototypes, and shipped features.
Mentor engineers on agent patterns, prompt engineering hygiene, evaluation discipline, and LLM failure modes.
Evaluate emerging models, frameworks, and production patterns.
Requirements:
8+ years of professional software engineering experience, with at least 2 years building production LLM or agentic systems (not just notebooks or demos).
Solid software engineering experience - front-end, APIs, async patterns, queues, databases, and the failure modes of distributed systems.
Demonstrated ownership of major features or subsystems in production.
Demonstrated experience mentoring junior engineers and raising team quality standards.
Demonstrated experience with event-driven systems: enrichment, retries, dead-lettering, backpressure.
Experience managing containerized applications in Kubernetes, EKS, ArgoCD, operators, Kustomize.
Deep, hands-on experience with at least one modern agent framework (LangGraph strongly preferred) and a track record of shipping agents that actually run, fail gracefully, and recover.
Real experience with evals — golden sets, offline and online evaluations, and using them to make ship/no-ship calls.
Production MLOps fluency: deploying LLM workloads under real latency, cost, and reliability constraints, and instrumenting what you ship.
Strong proficiency in Python; comfortable in TypeScript / Node.js for integrating with Worth’s services.
Clear, calibrated communicator — able to explain agent trade-offs to product, security, and customers without hand-waving.
Operates with extreme ownership in ambiguous, fast-moving environments.
Excited to work alongside a team that values “One Team”, “Extreme Ownership”, and “Create Raving Fans.”