Software Developer L2

Key Responsibilities

• Build and extend LangGraph-based agents (e.g., the Concierge agent, inventory and customer-affinity agents) with typed tool contracts, guardrail integration, and multi-tenant configuration.
• Develop MCP tool servers that wrap Aptos domain APIs (product catalog, customer profile, inventory, orders) with Pydantic v2 schemas, risk classification, and authorization checks.
• Implement Foundry platform features: tool cataloging and certification workflows, sandbox test environments, promotion to production, and secure connectivity into isolated client environments.
• Design and operate services on AWS EKS with Helm charts, Harness CI/CD and GitHub Actions pipelines, and the mandatory security gate (pip-audit, bandit, trivy, gitleaks).
• Implement event-driven data flows on Kafka and data services on Snowflake/Postgres; keep tenant configuration and caching layers correct under multi-tenant load.
• Instrument everything: OpenTelemetry traces into Langfuse, Prometheus metrics, structured logging; use traces to debug agent behavior and drive evaluation.
• Lead design reviews, mentor engineers, and contribute to shared platform libraries consumed by every SCP service.

Required Qualifications

• 3 to 6 years of software engineering experience, with strong production Python (FastAPI or similar async frameworks, Pydantic, pytest).
• Hands-on experience shipping LLM-powered features: agent frameworks (LangChain/LangGraph or equivalent), tool/function calling, prompt engineering, and provider APIs (Anthropic, OpenAI, Bedrock).
• Solid Kubernetes and Docker experience — deploying, debugging, and operating services on a managed cluster (EKS preferred), including Helm.
• Experience with at least two of: Kafka, Snowflake, Postgres, Redis, Couchbase, ClickHouse.
• Working knowledge of API security: OAuth2/JWT, mTLS, least-privilege authorization; comfort building in a security-gated CI/CD environment.
• Demonstrated ownership of services in production: on-call, incident response, observability-driven debugging.

Preferred Qualifications

• Experience building or consuming MCP servers, or designing tool interfaces for LLM agents.
• Familiarity with guardrails engineering (prompt-injection defense, PII redaction) or LLM evaluation/tracing tools (Langfuse, LangSmith).
• Experience with vLLM, GPU inference, or model serving infrastructure.
• Exposure to multi-tenant SaaS architecture or internal developer platforms; retail-domain experience a plus.